Brendan Eich: JavaScript, Firefox, Mozilla, and Brave

00:00:00.000 | The following is a conversation with Brendan Eich,

00:00:02.560 | creator of the JavaScript programming language,

00:00:05.680 | co-founder of Mozilla, which created the Firefox browser,

00:00:09.200 | and now co-founder and CEO of Brave Software,

00:00:12.840 | which has created the Brave browser.

00:00:15.440 | Each of these are revolutionary technologies.

00:00:18.320 | JavaScript is one of the most widely used

00:00:21.120 | and impactful programming languages in the world.

00:00:24.880 | Firefox pioneered many browser ideas that we love today,

00:00:29.000 | or even take for granted today.

00:00:31.280 | And Brave is looking to revolutionize

00:00:33.520 | not only the browser, but content creation online

00:00:36.640 | and the nature of the internet to make it fundamentally

00:00:39.640 | about respecting people's control over their data.

00:00:43.400 | Quick mention of our sponsors,

00:00:45.360 | the Jordan Harbinger Show,

00:00:47.360 | Sun Basket Meal Delivery Service,

00:00:49.720 | BetterHelp Online Therapy,

00:00:51.520 | and 8sleep Self-Cleaning Mattress.

00:00:53.920 | Click the sponsor links to get a discount

00:00:56.200 | and to support this podcast.

00:00:58.080 | As a side note, let me say that there's a tension

00:01:00.840 | between theory and engineering

00:01:03.160 | that I've been thinking a lot about.

00:01:05.200 | I tweeted something like,

00:01:06.960 | "Good execution is more important than a good idea,

00:01:09.880 | "but one helps the other."

00:01:11.800 | I think the wording of that sucks,

00:01:13.480 | but what I mean is a good idea is a must.

00:01:16.880 | But in my experience, good ideas are in abundance.

00:01:20.720 | Good execution, on the other hand, is rare.

00:01:23.720 | I think some mix of good timing, good idea,

00:01:26.040 | and good execution is essential.

00:01:28.120 | Getting that mix right is tough,

00:01:29.840 | and Brendan somehow, multiple times in his career,

00:01:32.880 | did just that.

00:01:34.240 | I'm starting to believe it's more art than science,

00:01:36.640 | like most interesting things in life.

00:01:39.200 | If you enjoy this thing, subscribe on YouTube,

00:01:41.680 | review on Apple Podcasts, follow on Spotify,

00:01:44.640 | support on Patreon, or connect with me on Twitter,

00:01:47.440 | @lexfriedman.

00:01:48.800 | And now, here's my conversation with Brendan Eich.

00:01:52.400 | When did you first fall in love with programming?

00:01:56.000 | - I didn't program a lot when I was in high school,

00:01:58.080 | but I had a friend who had a Commodore PET.

00:02:01.320 | And after we saw Star Wars, he said,

00:02:02.920 | "Hey, let's make a basic program

00:02:05.440 | "that does the Death Star Trench run."

00:02:07.800 | And it was just simple 2D graphics.

00:02:10.680 | And I didn't know what I was doing,

00:02:11.520 | so I just helped him out on the math and stuff like that.

00:02:14.040 | I was a math and science kid.

00:02:15.880 | I was really into the HP calculators of the early mid '70s.

00:02:19.840 | These were the RPM.

00:02:21.320 | They were really strongly built,

00:02:23.200 | and are at Goldfinger, instead of gold divinely heavy.

00:02:27.360 | There's probably some gold in them, too, gold metalization.

00:02:29.680 | But they were awesome calculators,

00:02:31.080 | and they had all the scientific functions,

00:02:32.960 | so I was really into that.

00:02:34.240 | So I aimed toward physics.

00:02:37.160 | I was a little late for the, I think,

00:02:40.440 | the 20th century golden age,

00:02:42.280 | and I read a lot of science fiction,

00:02:43.560 | so I was like, yeah, it's on the hyperdrives and warp drives.

00:02:46.480 | And physics was not gonna get there quickly,

00:02:50.160 | and I started hacking on computers

00:02:52.280 | while I was studying physics as an undergraduate

00:02:54.880 | at Santa Clara University.

00:02:56.400 | And I dodged the Fortran bullet

00:02:59.920 | because I was in the science department

00:03:01.840 | instead of the engineering department,

00:03:02.960 | where they still did Fortran card decks.

00:03:05.560 | I think they had an auto collator.

00:03:07.320 | But we were using Pascal,

00:03:09.160 | and I got one of the first portable C compilers,

00:03:13.640 | ports to the deck minicomputers we were using.

00:03:17.560 | And I fell in love with programming

00:03:19.320 | just based on procedural abstraction, Pascal,

00:03:24.320 | just what now would be considered old school,

00:03:28.880 | like structured programming from the '70s.

00:03:31.600 | Niklaus Wirth, the creator of Pascal,

00:03:33.480 | was a good writer and a good pedagogue.

00:03:36.160 | He always, at ETH, would do these courses

00:03:38.200 | where it's like build your own computer,

00:03:40.120 | build your own compiler, build your own operating system.

00:03:42.720 | - From scratch. - Yeah, kind of.

00:03:44.200 | And I know some people who are grad students under him

00:03:47.120 | said he would torture the students

00:03:50.480 | with things like this custom email system

00:03:52.240 | that had 25 word limit. (laughs)

00:03:55.120 | And things like that.

00:03:56.800 | - I unfortunately dodged both the Pascal

00:03:59.200 | and the Fortran bullets.

00:04:00.720 | Could you maybe linger on the Pascal?

00:04:04.760 | What kind of programming language was it?

00:04:06.280 | What is it reminiscent of today?

00:04:08.120 | Because it sounds like it may have had

00:04:10.840 | an impact on your own trajectory.

00:04:12.560 | - Yeah, it was in the Algol family.

00:04:14.440 | And Algol was the big successful

00:04:19.080 | language design and compiler project in the '60s.

00:04:23.120 | It had a successor called Algol 68,

00:04:24.920 | which was ambitious but not as successful.

00:04:27.400 | But Pascal was kind of a wordy procedures

00:04:30.960 | and functions language.

00:04:32.560 | It distinguished between functions,

00:04:33.880 | which return a value, and procedures, which don't,

00:04:36.120 | which just compute.

00:04:37.840 | And you could say that whole Algol family went into Ada.

00:04:42.560 | Pascal had a second life thanks to Borland

00:04:44.560 | with Turbo Pascal, which was hugely successful.

00:04:48.020 | I think in large part due to Anders Helsberg,

00:04:51.000 | who then went to Microsoft and did C# and .NET

00:04:53.840 | with his team there and has done really well.

00:04:55.520 | Doing TypeScript, Type JavaScript.

00:04:58.020 | So yeah, there's a lineage here.

00:05:00.800 | But I was also interested in C and Unix

00:05:03.800 | by the time I was an undergrad.

00:05:05.680 | Because people were bringing Unix up

00:05:07.840 | on all sorts of hardware.

00:05:09.200 | I had some friends who were doing

00:05:10.200 | their own wire-wrapped computers, 6820 maybe.

00:05:14.480 | And I was wire-wrapping for my engineering course,

00:05:19.160 | 1609 or something simpler, building a computer on a board.

00:05:22.480 | And I wanted to build a more ambitious one

00:05:24.740 | and port Unix to it, but I picked the wrong processor.

00:05:26.960 | I picked the National Semiconductor NS16032,

00:05:30.920 | which was this amazing CISC,

00:05:34.040 | Complex Instruction Set Computer.

00:05:37.200 | And not the reduced instruction set computers

00:05:39.360 | that were just being contemplated into the mid '80s.

00:05:43.760 | - And RISC ultimately won out.

00:05:45.440 | - RISC-I in some ways, it dissolved into,

00:05:48.380 | you have both now.

00:05:49.220 | You have these super scalar architectures

00:05:50.800 | where Intel has kept probably too much backward compatibility

00:05:55.240 | at the instruction level.

00:05:56.800 | But there's a front end that parses that

00:05:58.920 | into these wide internal instructions.

00:06:01.240 | So the very long instruction word research

00:06:04.440 | that was also interesting at the time

00:06:06.880 | kind of became the microarchitecture

00:06:08.560 | inside the backward compatible Intel.

00:06:11.440 | But I picked the National Semichip

00:06:13.160 | and it never got made successfully.

00:06:14.840 | It was full of bugs and I never could have brought it up.

00:06:17.800 | But I went on out of physics after three years

00:06:21.160 | into math computer science.

00:06:23.240 | And like I said, I did it because I saw,

00:06:25.740 | I was being sort of childlike and naive about physics.

00:06:28.420 | And I thought, meanwhile, the valley is go-go for computers.

00:06:32.120 | The Apple II, the PC, the Intel.

00:06:36.960 | 8086, 8088 based PC, the IBM.

00:06:41.960 | Gave Microsoft the future for a somewhat fishy deal.

00:06:46.800 | - So it was wide open in the computing space,

00:06:48.640 | but in physics, you were as optimistic about physics as--

00:06:52.160 | - No, I mean, I was one of three brothers

00:06:54.680 | who were all in the same grade.

00:06:56.060 | I have a twin and a younger brother

00:06:58.800 | who skipped second grade and was with us

00:07:01.440 | the whole time after that.

00:07:02.960 | And he went on.

00:07:04.460 | He actually studied under Kip Thorne at Caltech.

00:07:07.520 | But he also didn't, he ended up in software.

00:07:09.720 | (laughing)

00:07:12.340 | - Does it make you sad that theoretical physics,

00:07:16.100 | even with string theory, hasn't really had

00:07:18.640 | any foundational breakthroughs in the latter part

00:07:21.900 | of the 20th century?

00:07:23.320 | - Yeah, in fact, I'd say the problem is theory

00:07:25.220 | over experiments.

00:07:27.120 | I would say we need more Aristotle and less Plato.

00:07:30.840 | Mathematics is not all physical.

00:07:33.800 | There are lots of mathematics that cannot be realized

00:07:36.240 | as far as I know in this world.

00:07:38.340 | So to understand the world, you need to do experiments.

00:07:41.000 | You need to not just dream up inductive theories

00:07:45.160 | that could have lots of alternative theories

00:07:47.860 | competing with them, with no way to decide between them,

00:07:50.100 | except aesthetics, which is not a good guide, in my opinion.

00:07:53.000 | - I don't know if you are friends with

00:07:54.920 | or have a relationship with Elon Musk.

00:07:56.860 | Where's the, in terms of what you would love to see

00:08:00.500 | our society investing in building up,

00:08:02.760 | is it closer to Elon or is it closer to

00:08:06.320 | Feynman and Einstein and those?

00:08:09.120 | - Well, those gentlemen are no longer with us,

00:08:10.560 | and I think that's noticed.

00:08:11.960 | So like I said, the real glory days of physics,

00:08:14.440 | the famous pictures from Germany before the Second War

00:08:17.240 | were just a fantastic assembly of brains,

00:08:21.720 | Schrodinger and Einstein.

00:08:23.680 | Physics, I think, took a wrong turn

00:08:26.280 | that maybe all of, I would say, Western science took

00:08:30.320 | in going for models over reality, right?

00:08:33.720 | You see this in all sorts of fields.

00:08:35.280 | Now, we can build models that are very predictive

00:08:37.980 | and generative and then we build actual devices

00:08:40.280 | or semiconductors, things like that.

00:08:42.160 | That's good, I'm not dismissing that.

00:08:44.400 | We need good models, we need to experiment

00:08:46.840 | and prove them and test them.

00:08:49.640 | But the problem I've seen in physics,

00:08:51.880 | which you see certainly in economics, the dismal science,

00:08:55.600 | and you see, surprisingly, in other so-called hard sciences,

00:08:59.040 | is models that don't really have to be tested

00:09:04.040 | against reality.

00:09:05.520 | They can instead become policy tools

00:09:07.620 | or they can become, like I said,

00:09:09.880 | one of a large family of alternate theories

00:09:12.920 | that could be as predictive,

00:09:14.600 | but nobody's doing the winnowing out.

00:09:16.960 | - That's such an interesting tension in society.

00:09:18.840 | You see this in even the softer sciences,

00:09:20.960 | which I have a deep love for, like psychology.

00:09:24.400 | You see this in epidemiology, not with the virus.

00:09:27.200 | - Absolutely.

00:09:28.040 | - There's this tension of how much of the world

00:09:31.280 | can we understand through just a beautifully fit model?

00:09:35.060 | And then at the same time, my main work

00:09:37.640 | is in machine learning, where it's like

00:09:41.000 | there is no provable thing usually.

00:09:44.400 | It's just kind of, it's all about just getting

00:09:46.880 | the right data set and getting tricks and so on.

00:09:49.640 | And there's this tension, even in my own soul,

00:09:51.760 | of like I grew up in theoretical computer science.

00:09:55.680 | I loved approximation algorithms,

00:09:58.680 | like all of that, like different complexity classes,

00:10:01.940 | just those little puzzles.

00:10:03.960 | I mean, I don't know, to you as somebody

00:10:06.200 | who was in math and computer science

00:10:08.600 | and then ended up going into places

00:10:11.280 | where you engineered some of the most impactful things

00:10:13.860 | in this world, do you see the P versus NP,

00:10:18.680 | all that whole space as interesting at all?

00:10:21.280 | - Yeah, it's not that useful in practice, right?

00:10:24.640 | People are using it with sort of crypto analysis

00:10:27.920 | or asymptotic arguments about can we have

00:10:31.760 | a quantum-resistant crypto algorithm, things like that,

00:10:35.320 | which may not be practical, right?

00:10:36.600 | If you follow Mikhail Diakonov or Gil Kalai,

00:10:40.160 | there are big questions about how quantum computing

00:10:43.640 | will scale up, how practical it will be.

00:10:46.800 | - Is that something that you think about,

00:10:48.000 | quantum computing?

00:10:49.480 | - Not for a spare time.

00:10:51.080 | Like you said, I'm not using this kind of computer science

00:10:53.560 | in practice because almost everything now is engineering

00:10:57.680 | and finding ways to get computers

00:11:02.160 | to be more useful for people,

00:11:03.340 | which goes from design problems,

00:11:06.160 | which are really kind of an art.

00:11:07.440 | Like Knuth said, anything you can't automate is an art.

00:11:09.960 | Well, we can have machine learning compose music

00:11:13.320 | and it can imitate, you can train it,

00:11:14.800 | and it can sound kind of decent,

00:11:16.040 | but maybe lacking that je ne sais quoi.

00:11:18.680 | But user interface still, I think, requires human art.

00:11:23.560 | - So speaking of things that didn't follow

00:11:26.440 | a perfect theory and model, JavaScript.

00:11:30.480 | So there's two things.

00:11:31.320 | One, it had an impact on the world at a huge scale,

00:11:34.360 | obviously, and it also still is one of,

00:11:38.320 | probably the most popular programming language in the world.

00:11:41.400 | So can we go back to the origin story?

00:11:44.140 | Can you tell the story of how JavaScript was created?

00:11:47.840 | - Yeah, I was at Silicon Graphics after graduate school

00:11:51.080 | for seven years, and it got to be big and successful

00:11:54.720 | and divisionalized and political,

00:11:56.800 | and I thought, kind of boring.

00:12:00.480 | And a friend who'd been there went to

00:12:03.240 | one of the last of the super companies,

00:12:05.280 | super startups in the early '90s.

00:12:07.600 | There were several.

00:12:08.720 | I suppose General Magic was a little after that

00:12:10.840 | or around the same time.

00:12:11.920 | But MicroUnity was that company that I went to,

00:12:14.360 | and it was because my friend, Jeff Weinstein,

00:12:16.880 | had gone there from Silicon Graphics.

00:12:18.040 | He recruited me, and MicroUnity was doing everything.

00:12:21.960 | So this was like the ultimate sort of pretend grad school.

00:12:26.080 | It was doing a new fab, new semiconductor process.

00:12:29.360 | It was doing new analog and digital circuits

00:12:32.600 | on the same very large but not wafer-scale chip.

00:12:36.080 | Originally, it was five centimeters on a side.

00:12:40.360 | It was really hot, too, so it needed a water cooler.

00:12:43.880 | It was a Craykiller, and then they shrunk it,

00:12:46.280 | and they tried to do a home sort of media processor

00:12:49.320 | that was essentially a barrel processor,

00:12:51.920 | but you could think of trying to do all the things

00:12:55.440 | that we now see in modern architectures

00:12:58.160 | with short vector instructions

00:13:00.240 | and sort of wide instructions for multiple issue,

00:13:04.640 | and doing a lot of the stuff in software

00:13:07.520 | because the second iteration, the set-top box,

00:13:10.120 | was really for avoiding the cost to the cable company

00:13:13.040 | of rolling the trucks out

00:13:14.520 | to replace your garbage General Atlantic set-top box

00:13:17.600 | with a totally newer, less garbagey one.

00:13:21.040 | So if you could have software-gradeable set-top boxes,

00:13:23.900 | the cable companies thought they could save a lot of money

00:13:26.240 | and add features.

00:13:27.200 | - Is this assembly, or which level of the software?

00:13:30.280 | - It was like, we were writing in, we were using GCC.

00:13:33.840 | We were writing C++ in C.

00:13:36.280 | Somebody I worked with there, really very smart guy,

00:13:39.920 | hired from a sort of Wall Street

00:13:43.040 | hotshot programming consultancy,

00:13:45.240 | did his own hardware design as well as software.

00:13:47.680 | We were working on how to make

00:13:49.400 | not only short vector units,

00:13:50.740 | but general bit shufflers and permuters

00:13:53.080 | so you could do things like crypto algorithms efficiently,

00:13:57.480 | and you could do demodulation of the cable,

00:14:01.400 | complex quadrature amplitude modulated signal.

00:14:05.240 | So you're basically taking A to D converters,

00:14:07.980 | dumping things in buffers,

00:14:09.000 | and then doing the rest in software.

00:14:10.760 | All the framing and the Reed Solomon

00:14:12.880 | and Viterbi and all that error correction.

00:14:14.520 | So that was really great learning experience,

00:14:16.100 | but it was not gonna work.

00:14:16.940 | It was doing too many risky things at once.

00:14:19.440 | As Jim Clark said to me when I hopped to Netscape

00:14:21.960 | after three years at MicroUnity,

00:14:23.240 | he said, "Oh yeah, you do 10 things,

00:14:25.240 | "each one in 10 odds, it's gonna be one in 10 billion."

00:14:28.040 | (laughing)

00:14:29.880 | The multiplication principle.

00:14:31.080 | So Netscape was already a rocket,

00:14:33.360 | and I passed the chance to go there in 1994.

00:14:36.800 | I knew the founders, 'cause I worked at SGI,

00:14:39.240 | Clark's company.

00:14:40.080 | - Could you pause for a second?

00:14:41.160 | Netscape, when was the launch of this rocket?

00:14:44.800 | - '94.

00:14:45.640 | - '94 was the launch of Netscape?

00:14:47.280 | - And I went there in early '95, in April.

00:14:49.720 | - Okay, so you said you missed the launch.

00:14:52.160 | - Well, I missed the first floor employment opportunity,

00:14:55.560 | but the IPO was August 1995, so I was there for that.

00:14:59.000 | - How obvious was it that Netscape was world changing?

00:15:02.280 | What was the layout?

00:15:04.080 | Was Netscape one of the first big browsers?

00:15:06.160 | - Yes, so when I was at MicroUnity still in '93,

00:15:09.280 | we saw a browser called Mosaic.

00:15:11.720 | And up 'til then, we'd used email,

00:15:13.400 | and we'd used Usenet, the NNTP protocol.

00:15:15.840 | We'd used Newsreaders, we used FTP,

00:15:18.680 | we used all these old internet protocols,

00:15:20.880 | all relying on the DNS and TCP/IP and UDP,

00:15:24.080 | for that matter.

00:15:25.720 | When I was at Silicon Graphics,

00:15:26.800 | we brought up the whole stack, right?

00:15:28.120 | We had to discover how to find the ethernet addresses

00:15:31.760 | on your network, and then find IP addresses for them,

00:15:34.720 | ARP protocol, all that stuff.

00:15:35.960 | And it was great, because nobody knew in the '80s

00:15:38.680 | what was gonna win.

00:15:39.520 | All the proprietary stacks like IBM, SNA, and DECnet,

00:15:43.840 | and all these other protocols were saying,

00:15:45.720 | we're gonna do it, or it's gonna be heterogeneous future.

00:15:48.680 | Instead, it was Berkeley Unix and the TCP/IP stack

00:15:52.560 | that dated back to the ARPANET that won.

00:15:54.960 | And I think we knew it.

00:15:56.480 | We all knew it at SGI, but the salespeople didn't.

00:15:59.320 | And so they kept trying to get multiple network stacks

00:16:02.120 | interoperating, but in the end, it won.

00:16:04.400 | And so that was the internet.

00:16:06.520 | And it was email and texty, and it was used then,

00:16:09.000 | very texty.

00:16:09.840 | And then Tim Berners-Lee did his thing,

00:16:12.580 | but I don't think I was paying attention.

00:16:14.160 | And I think the date when he first did it,

00:16:16.440 | or when he wrote the famous email,

00:16:17.840 | it's been pushed back to '89.

00:16:19.620 | But I noticed a mosaic in '93,

00:16:21.720 | because one of the things that Mark Andreessen

00:16:24.480 | and Eric Bina did at NCSA was they innovated

00:16:28.240 | on the early HTML standard.

00:16:31.360 | They, in particular, Mark sent this email saying,

00:16:33.080 | hey, everybody, we think you should be able

00:16:34.400 | to put an image in a page.

00:16:36.280 | And you know when he sent that,

00:16:37.640 | Eric Bina had already written the code.

00:16:39.080 | And I talked to Tim Berners-Lee more recently,

00:16:41.920 | just a few years ago, and he was like,

00:16:43.120 | oh, we had another way of doing it.

00:16:44.280 | And it didn't work out, 'cause Mark shipped his in mosaic.

00:16:47.960 | And this convinced me of several things.

00:16:50.040 | One, the internet meant there was a huge

00:16:53.160 | first mover advantage, and being fast,

00:16:56.440 | getting on first, mattered a lot.

00:16:58.160 | And so Richard Gabriel of Scheme and Poetry fame

00:17:01.880 | has written about this, the famous essay.

00:17:03.080 | - Poetry, what's poetry?

00:17:04.760 | - Well, he's a poet.

00:17:06.000 | - Oh, actual poetry.

00:17:07.160 | He's like talking about some kind of subject.

00:17:08.320 | - No, no, I mean, he's the founder of Lucid,

00:17:10.160 | which is where Jamie Zawinski worked before Netscape.

00:17:12.840 | And Lucid was doing compilers, and Lucid Emacs,

00:17:16.280 | which was a fork of Emacs, famously Jamie fighting

00:17:19.480 | against Richard Stallman, Stallmax.

00:17:22.640 | And so Richard Gabriel, very brainy computer guy,

00:17:25.240 | but also a poet, but he wrote a nice essay

00:17:27.320 | that gets abused all the time.

00:17:28.920 | In fact, Jamie's put a kind of warning

00:17:30.440 | in front of his version of it on his site, JWZ.org,

00:17:33.320 | called Worse is Better.

00:17:34.360 | And this is about survival advantage of software

00:17:38.240 | in the network world, in my opinion.

00:17:39.760 | It's about Unix, it started out being framed

00:17:42.640 | as Unix and Lisp, good news, bad news,

00:17:44.600 | because all the Lisp people, the MIT people,

00:17:46.560 | were like, oh, the crown jewel scheme,

00:17:50.320 | this Fabergé egg, or Common Lisp, this giant cathedral.

00:17:53.760 | Of course we're gonna win, this is civilization.

00:17:55.680 | And those farmers in New Jersey, to borrow from Sopranos,

00:17:59.080 | those hicks down at Bell Labs, they're just,

00:18:02.440 | there's nothing sound there, it's all hacking.

00:18:04.160 | - Yeah.

00:18:05.240 | - Well, guess what won?

00:18:06.800 | - Wow, so you're saying this is a fundamental principle

00:18:11.080 | of the internet is moving fast winds.

00:18:14.760 | - You could say in almost any network system,

00:18:16.920 | like in biological evolution, you see successful alleles

00:18:19.760 | sweep populations, and they don't always have,

00:18:22.000 | they aren't free of flaws, they,

00:18:24.840 | heterozygous advantage, right?

00:18:26.320 | You can get, both parents give you the gene variant

00:18:30.160 | and you get sickle cell anemia, right?

00:18:31.760 | But if one of them does, you're more resistant to malaria.

00:18:34.640 | And so this isn't a beautiful process,

00:18:38.120 | except at large scale, and then you realize that

00:18:40.440 | because it moves fast and can adapt, it can win.

00:18:44.200 | And people still struggle with this.

00:18:46.560 | I used to struggle with this, because JavaScript was done

00:18:48.800 | in such a hurry, and the force of web compatibility

00:18:52.040 | meant early mistakes couldn't be fixed.

00:18:55.200 | And even the standards process injected new mistakes,

00:18:57.640 | as it will, but often standards bodies go back

00:19:00.360 | and making compatible changes.

00:19:01.640 | You can't do that with the web.

00:19:03.040 | It's more like, again, like biology,

00:19:05.600 | you preserve what still works.

00:19:07.080 | You don't wanna break ATP metabolism or whatever.

00:19:10.320 | So you have to kind of resign yourself to the reality

00:19:15.120 | of worse is better being enshrined

00:19:19.960 | in actual design points you might not like.

00:19:23.080 | And that happened with JavaScript, and I'm way over it,

00:19:25.760 | but it also, I think, was a huge advantage,

00:19:28.720 | which is why JavaScript has kind of swept

00:19:31.160 | a lot of programming domains.

00:19:33.400 | People will say, "Oh, it's not because of merit."

00:19:34.840 | Well, you're right, but we also improved it over time

00:19:36.880 | in the standards body.

00:19:37.720 | I spent 20 years doing that.

00:19:39.320 | And you don't get that choice.

00:19:41.760 | It's like, I'm not saying that that was the best language.

00:19:45.360 | I'm just saying that was the right time to do it.

00:19:47.400 | And I like to say the alternative was not to do it.

00:19:50.280 | I could have told Netscape, "I can't do this.

00:19:52.120 | "It's too rushed."

00:19:53.480 | And it would have been Visual Basic Script,

00:19:55.240 | and it would have been bad.

00:19:58.040 | - So that's a good way to present the alternative.

00:19:59.800 | But so it was a Netscape, and you have written it

00:20:03.040 | in how many days, and why was it only that many days,

00:20:05.880 | and what was the goal and the underlying principles

00:20:08.880 | in your mind at the time?

00:20:09.720 | - So the whole, I'm sort of describing worse is better

00:20:11.760 | in a frenetic way because it fit the model of Netscape.

00:20:15.000 | When it was known that Jim Clark and Mark Inveas

00:20:19.000 | were founding Netscape, and they did the first release

00:20:21.400 | in 1994, that browser took over from Mosaic.

00:20:24.840 | In fact, that's why Mozilla is called that.

00:20:27.120 | It's the Mosaic killer.

00:20:28.240 | It's like the giant monster that kills Mosaic.

00:20:30.120 | - That's awesome.

00:20:30.960 | - And they knew that, it wasn't that, again,

00:20:34.160 | it's not like you're doing advanced scientific research

00:20:36.200 | that is changing the world.

00:20:37.360 | You're more like taking down the last iteration

00:20:41.000 | of the browser Mark did, which had images,

00:20:43.160 | and other importances before he stopped working on it,

00:20:45.880 | and you're making Netscape the new thing that has images,

00:20:49.040 | plugins, which was the way to do video back in the day.

00:20:51.680 | It had something that's kind of died now

00:20:53.200 | for tiled windows called frames and framesets.

00:20:56.880 | - Oh, yeah, yeah, yeah, yeah.

00:20:57.920 | - HTML tables, that was new.

00:20:59.280 | Eric Bina did tables in Netscape 1.1.

00:21:01.640 | So when I got there, they were heading toward IPO.

00:21:05.040 | Clark wanted to IPO early.

00:21:06.480 | I think his instinct was right,

00:21:07.640 | and that kicked off the whole dot-com era.

00:21:10.280 | There was a recession in the US in '91.

00:21:13.280 | You can see old Law & Order reruns

00:21:15.140 | where they talk about the recession,

00:21:16.520 | how hard it's hitting New Yorkers.

00:21:18.080 | And after that, Greenspan really goosed things

00:21:20.440 | at the Federal Reserve, and technology had been

00:21:23.160 | sort of fermenting in a way that came together

00:21:25.800 | with the internet, and Netscape made it possible

00:21:28.720 | to do pets.com, to do eBay, to get people to recognize

00:21:33.160 | a URL on a billboard and then type it in when they get home.

00:21:36.040 | And that was huge.

00:21:37.320 | That was so fast moving a rocket that Mark

00:21:42.120 | and the engineering team there thought,

00:21:45.080 | we need to make this a programmable browser,

00:21:47.920 | not just a document viewer, not just a video.

00:21:50.840 | - So it was all HTML with images and tables,

00:21:53.800 | and also, like you said, frames.

00:21:55.280 | - Early plugins.

00:21:56.120 | - There's no dynamic element at all.

00:21:57.800 | - Yeah, the most dynamism you get was from a plugin,

00:21:59.680 | which there were a few of them then.

00:22:01.640 | Flash didn't exist at that point.

00:22:03.960 | It was, I think--

00:22:05.000 | - Java Applets yet, or no?

00:22:07.400 | - Well, that's the thing.

00:22:08.240 | We did the deal with Sun.

00:22:09.440 | In fact, I was recruited to go do Scheme in the browser.

00:22:12.560 | Remember Guy Steele and Gerald Sussman's

00:22:14.800 | beautiful Lisp variant?

00:22:16.760 | I was gonna do it in the browser

00:22:17.920 | because my friends from SGI thought,

00:22:19.480 | hey, we like Scheme, you like Scheme.

00:22:20.920 | And I'm like, I hardly ever use Scheme.

00:22:22.140 | It's not really used in industry,

00:22:23.400 | except in sort of silos.

00:22:25.000 | But I like it.

00:22:27.080 | Okay, I'll come do Scheme in the browser.

00:22:29.760 | I have a slide from my 2017 talk

00:22:32.720 | where I have Bruce Willis crawling through the duct

00:22:34.640 | in Die Hard.

00:22:35.480 | He's like, come out to the coast, have a lot of fun.

00:22:38.280 | Come on, do Scheme in the browser.

00:22:40.560 | But when I got there, there was no Scheme in the browser

00:22:42.160 | because they'd started a deal with Sun Microsystems.

00:22:45.240 | And my best contact there was Bill Joy,

00:22:47.040 | who I admired as a Berkeley Unix founder and Sun founder.

00:22:50.560 | And Bill got the idea of making the browser programmable too.

00:22:53.880 | And so the main idea was to put the Java VM,

00:22:57.200 | which at that point was not really easy to embed,

00:23:00.360 | into Netscape, including the Netscape version on Windows

00:23:04.220 | that was still most popular,

00:23:05.600 | which was the 16-bit Windows 3.1, which was going away.

00:23:09.520 | Microsoft was coming out with Windows 95

00:23:11.320 | and everyone was afraid they were gonna do

00:23:13.560 | Internet Explorer, I guess, two at that point,

00:23:16.160 | three the next year.

00:23:17.040 | They already bought or invested in somehow Spyglass,

00:23:20.920 | this other company that shot out from NCSA

00:23:24.440 | at University of Illinois.

00:23:26.360 | And in fact, Microsoft had tried to buy Netscape

00:23:29.720 | in late '94, before I got there.

00:23:31.360 | And I heard about this later.

00:23:32.240 | I heard they offered way too little money.

00:23:35.000 | And so Jim Barksdale and Jim Clark said,

00:23:37.360 | "Get out of here," in a pound of sand.

00:23:39.760 | But then they realized, oh, this is going to hurt us

00:23:43.740 | because now they're gonna copy us.

00:23:45.480 | Didn't happen right away.

00:23:46.320 | I'm not sure when Gates' Internet Tidal Wave memo

00:23:49.080 | was written.

00:23:50.240 | That's the famous memo he wrote when Bill Gates realized

00:23:53.160 | that Microsoft was going down this old copy AOL path

00:23:56.840 | or copy CompuServe path, a project called Blackbird,

00:24:01.100 | presumably after the SR-71, I don't know.

00:24:03.240 | But they were gonna make a dial-up service

00:24:05.360 | with a custom content language stack and custom rendering.

00:24:09.040 | It wasn't the web.

00:24:10.000 | They could have content partners.

00:24:13.200 | They have a lot of money.

00:24:14.400 | But it still wasn't to scale the web.

00:24:15.920 | It wasn't gonna be compelling.

00:24:16.880 | And Gates realized this, and he turned the company

00:24:19.480 | on a dime, and they couldn't buy Netscape.

00:24:22.400 | Again, I'm not sure of the timing.

00:24:23.320 | So they decided to copy it.

00:24:24.540 | And once we realized that, everybody inside Netscape

00:24:27.400 | felt even more urgency and more of a frenetic mood.

00:24:30.920 | And so my chance to do Scheme disappeared

00:24:33.620 | when the Java deal started brewing.

00:24:35.840 | But there was still a chance to do a companion language

00:24:38.840 | to Java because Java is a compiled language.

00:24:42.620 | It's evolved and improved quite a lot since then too.

00:24:44.960 | But it was for serious advanced programmers

00:24:48.560 | that cost a certain salary or hourly rate.

00:24:51.400 | And people observed, Bill Joy observed,

00:24:53.760 | and Mark Andreessen and I observed,

00:24:55.760 | that in a mature stack like Microsoft,

00:24:57.900 | you really benefit from having a scripting language

00:25:00.020 | like Visual Basic, which became Visual Basic Script

00:25:03.000 | in IE3, but didn't take over and kill JavaScript,

00:25:07.120 | that you need two languages.

00:25:08.360 | One is for the component writers

00:25:10.160 | who are higher priced and more expert.

00:25:12.560 | And the other is for scripters,

00:25:17.000 | certified public accountants, designers,

00:25:19.480 | graphic designers with some programming inclination.

00:25:21.720 | Anybody, amateurs, doesn't matter.

00:25:23.460 | There's a much more demotic approach there

00:25:26.600 | for programming the components together,

00:25:28.640 | gluing them together.

00:25:29.880 | Some people say duct tape language, which I don't really like.

00:25:32.800 | But we saw, Bill Joy and Mark Andreessen and I,

00:25:36.080 | we saw the need for a companion language.

00:25:38.040 | And the gleam in our eye was to call it JavaScript.

00:25:40.400 | I didn't like it, that was marketing's plan.

00:25:42.480 | Mark called it Mocha, which I liked.

00:25:45.080 | And Netscape Marketing, I think, didn't like that.

00:25:47.080 | So they said, "Oh, there's some trademark

00:25:48.440 | "in some software somewhere that uses Mocha,

00:25:50.740 | "so we can't use that."

00:25:52.120 | And they tried LiveScript in August and that didn't last.

00:25:54.960 | And then finally, we got the trademark license

00:25:56.680 | in December, 1995.

00:25:58.720 | But the work I did to prove that it could be done

00:26:01.620 | was important because I came in in April

00:26:05.200 | and even then Netscape was growing so fast

00:26:08.320 | that they couldn't find an open hiring requisition

00:26:11.280 | in the client team for me.

00:26:12.320 | So they hired me into the server team.

00:26:14.280 | And I worked for a month on server team

00:26:16.680 | on what became HTTP/1/1.

00:26:18.600 | So I was actually, I had done protocol work

00:26:20.280 | at Silicon Graphics with Greg Chesson,

00:26:22.640 | former Bell Labs intern, grad student intern,

00:26:25.360 | who knew all the Unix founders.

00:26:27.040 | And Greg was very interested in taking protocols

00:26:30.920 | to the next level with VLSI

00:26:32.840 | 'cause he thought that CPUs wouldn't scale up.

00:26:35.480 | He was mistaken in that, unfortunately.

00:26:37.480 | Moore's law more than kept up.

00:26:39.060 | And you have gigabit ethernet running

00:26:41.400 | with conventional processors.

00:26:42.520 | But I worked on protocols at SGI

00:26:45.680 | as well as Unix kernel hacking and NFS and things like that.

00:26:49.000 | So I came into Netscape to work on the server side

00:26:53.060 | for a month, but I was thinking the whole time,

00:26:55.280 | what should this language be like?

00:26:56.360 | Should it be easy to use?

00:26:58.480 | Might it syntax even be more like natural language

00:27:00.680 | like HyperTalk, which is Bill Atkinson's language

00:27:04.840 | in HyperCard, if you have ever used HyperCard

00:27:07.320 | on an early Mac.

00:27:08.800 | And I thought, well, I'd like to do that,

00:27:12.080 | but my management is saying make it look like Java,

00:27:15.160 | which looks like C from a distance.

00:27:17.480 | - What does that mean?

00:27:18.480 | Is it braces we're talking about visually?

00:27:20.400 | Does it mean like, what, management,

00:27:23.040 | do they understand what--

00:27:24.160 | - They don't, marketing didn't know,

00:27:25.520 | but management did, like the rickshaw of engineering knew.

00:27:28.460 | And we had a plan even that was,

00:27:31.720 | if you have this companion language,

00:27:32.920 | you're going to glue things together

00:27:34.560 | between Java and JavaScript.

00:27:35.720 | So you're going to have commerce in memory,

00:27:38.200 | in the heap with data types.

00:27:40.240 | So you're going to want some of the data types in Java

00:27:42.760 | to reflect in the JavaScript.

00:27:43.880 | You're going to want the primitive types that Java,

00:27:46.120 | unfortunately, severed from objects.

00:27:47.520 | So at least some of them, double, let's call it

00:27:50.880 | in Java's terms, from the C term for double precision,

00:27:53.720 | floating point, or strings, or Booleans, and objects.

00:27:58.720 | And so right away, there was this constraint

00:28:03.960 | that looking like Java meant kind of the C curly brace

00:28:07.520 | syntax, but also some of the data types and objects.

00:28:09.840 | - Like objects and so on, all that kind of stuff.

00:28:11.520 | - Dotted columns.

00:28:12.360 | - Comparison operator.

00:28:13.200 | - Garbage collection, all that stuff.

00:28:15.480 | Even the bitwise operators and the shift operators,

00:28:17.640 | including the unsigned right shift,

00:28:19.580 | which Java had because it didn't have unsigned integer types.

00:28:22.720 | It said if you want to do unsigned operations,

00:28:24.600 | use an operator.

00:28:25.440 | And that turned out to be important much later.

00:28:27.320 | I'll tell that story five times.

00:28:28.560 | But JavaScript inherited a set of,

00:28:33.980 | operators, the expression grammar,

00:28:36.340 | the statement grammar up to a point from Java.

00:28:39.380 | But I wanted a functional language.

00:28:40.740 | I wanted a little bit of Scheme,

00:28:42.840 | even though it wasn't as clean as Scheme.

00:28:44.620 | - So you had a love, sorry to interrupt,

00:28:46.140 | you had a love for Scheme and Lisp,

00:28:47.660 | that functional language landscape.

00:28:49.740 | - Yes, I wanted first class functions.

00:28:51.620 | Because I saw the need for callbacks in the browser,

00:28:54.000 | where it's a single threaded program.

00:28:55.940 | All the early browsers were single threaded.

00:28:57.380 | And it's the right model for users.

00:28:58.640 | Most users weren't ready for mutual exclusion

00:29:00.620 | and threading.

00:29:01.680 | So in a single threaded world,

00:29:03.660 | you cannot block the user interface.

00:29:05.420 | So you have to use a callback and run later.

00:29:07.500 | And without getting too fancy

00:29:09.900 | and trying to capture the continuation,

00:29:11.620 | like call CC does in Scheme,

00:29:13.340 | I thought I'll just make it easy to have fun arcs,

00:29:16.620 | first class functions you pass downward

00:29:18.860 | and it can be called back.

00:29:20.660 | And Java didn't have that at the time.

00:29:23.660 | It took forever to get proper first class functions,

00:29:27.300 | lambdas now into Java, Java seven or eight, I think.

00:29:30.740 | - It did have concurrency, right?

00:29:32.120 | - Yes.

00:29:32.960 | - From the very beginning.

00:29:33.780 | But you were thinking that the JavaScript in the browser

00:29:36.200 | would not have the luxury of being concurrent.

00:29:39.400 | - That's right.

00:29:40.240 | And the reason was,

00:29:41.080 | Java was gonna run in the plugins.

00:29:42.560 | We could fork threads and go to town.

00:29:44.600 | But the main action in the browser

00:29:46.920 | was in the single threaded program,

00:29:48.280 | in the single Unix process on Unix or Windows.

00:29:51.920 | And it was where you had to service the event loop

00:29:54.440 | and then go do things.

00:29:56.180 | Respond to the network, lay out some HTML,

00:29:58.600 | render it, turn widths into heights

00:30:01.100 | by filling containers, boxes,

00:30:03.180 | the early, what became the CSS box model.

00:30:06.060 | And run scripts to make the thing livelier,

00:30:09.580 | respond to user input.

00:30:11.340 | And all that event driven programming

00:30:14.180 | was in part like HyperCard

00:30:16.060 | because HyperCard had this on event name syntax.

00:30:19.420 | And so that's why you have in JavaScript on click

00:30:21.640 | run together as the name of the event handler.

00:30:24.080 | And there's some funny ones on mouse over and on mouse out.

00:30:27.060 | People still complain about those.

00:30:28.180 | But there are many more events now

00:30:30.460 | over the years standardized.

00:30:31.580 | But it was a mix of event driven,

00:30:33.940 | single threaded programming

00:30:34.900 | because it had to run in the main thread of the browser

00:30:38.140 | where the action is.

00:30:38.980 | And Java never got there,

00:30:40.660 | which meant Java could not interact easily

00:30:42.620 | or quickly or in a nested way with the document,

00:30:46.680 | with the objects reflected from the HTML document,

00:30:49.140 | with the tables and forms and so on.

00:30:51.180 | And that is one of the reasons I think JavaScript survived

00:30:54.460 | and Java kind of died.

00:30:55.300 | Java was in this plugin prison.

00:30:56.900 | It essentially was confined to a rectangle,

00:30:58.700 | the app with rectangle.

00:30:59.860 | And while we even built a next year,

00:31:02.660 | Nick Thompson, a friend from SGI,

00:31:04.820 | who was an intern grad student at CMU at the time,

00:31:08.580 | built the first version of Live Connect

00:31:10.220 | to glue Java and JavaScript together

00:31:12.580 | to deliver on that vision where you do have commerce

00:31:14.620 | between the data types in the heap.

00:31:16.780 | - Did it work?

00:31:18.820 | - It worked, but Java was in charge.

00:31:21.220 | JavaScript was in charge

00:31:22.180 | and Java was just these components, these helper.

00:31:24.380 | - You might as well do everything in JavaScript.

00:31:25.900 | - What happened over time.

00:31:26.740 | It's like an evolutionary filter.

00:31:27.940 | It just kind of, who needs the plugin?

00:31:29.580 | And in fact, Sun mismanaged Java as a plugin.

00:31:33.020 | They thought, oh, Netscape was giving us

00:31:35.500 | the distribution vehicle and we don't care

00:31:37.700 | about the browser, it's just about getting Java out there.

00:31:39.700 | And that was a big miscalculation.

00:31:41.220 | They then tried, because Microsoft's killing Netscape

00:31:43.820 | after a few years, they tried getting into Microsoft.

00:31:46.140 | And you may remember there was a Sun-Microsoft deal

00:31:48.500 | which famously blew up.

00:31:50.660 | And Microsoft kicked Java out of Windows.

00:31:54.820 | And that's when they really pulled the trigger.

00:31:56.420 | I think they already evaluated it and liked it

00:31:58.140 | on Anders Helsberg's .NET and C#

00:32:01.740 | and decided we're gonna just not have Java.

00:32:04.620 | We don't want any of that Sun stuff.

00:32:06.460 | We don't want the patent risk.

00:32:07.540 | We don't want, I'm not sure what the fights were about.

00:32:10.100 | There was some patent angle to it, I think.

00:32:12.300 | And up till then, Microsoft had been using Java components

00:32:16.940 | like in Outlook Web Access, which had a lot of JavaScript

00:32:20.820 | to be a Hotmail-like user interface.

00:32:23.100 | They had to call the mail server through HTTP

00:32:27.740 | and they used a Java object to do this.

00:32:31.340 | And when they gave the boot to Sun,

00:32:33.820 | they suddenly other, the left-hand gave the boot

00:32:37.260 | and the right-hand said, "We better do something else

00:32:39.140 | in Outlook Web Access.

00:32:40.060 | What are we gonna do?"

00:32:40.900 | And they said, "Let's just add an ActiveX component,"

00:32:43.620 | which is their own native way

00:32:45.380 | of embedding things in languages.

00:32:46.900 | And we'll make it, it'll be what became XML HTTP request,

00:32:51.980 | which is now a web standard for calling asynchronously.

00:32:55.100 | And it's been replaced by the Fetch API

00:32:57.060 | in HTML5 or HTML living document.

00:32:59.500 | But this whole lineage goes back

00:33:01.100 | to Java being successfully the loser and getting kicked out.

00:33:05.100 | And after Microsoft kicked it out, it was a plugin

00:33:07.680 | and you would find it required for like smart card banking

00:33:11.020 | in the Nordic countries where that was mandated by law

00:33:14.700 | but really didn't get used much.

00:33:16.380 | Or there were pilots who used it for flight information.

00:33:21.560 | But Flash, which Netscape could have bought

00:33:24.180 | but unfortunately didn't.

00:33:25.540 | - Unfortunately didn't?

00:33:26.380 | - Yeah, we would have screwed it up.

00:33:28.100 | - What year are we talking about with Flash?

00:33:29.540 | - I think after the IPO, so it was probably late '95.

00:33:32.180 | - Flash was around, was it Adobe?

00:33:34.140 | No, it wasn't.

00:33:34.980 | - No, it was called Future Splash

00:33:36.380 | and it was these brothers, Jonathan Gay,

00:33:38.700 | I think his name was.

00:33:39.540 | He came knocking and the marketing guy at Netscape

00:33:43.540 | who was screening the technology partners

00:33:45.820 | or wannabe acquisitions was brutal

00:33:47.780 | and just everybody wanted to get in on the Netscape

00:33:50.300 | stock gravy train and he sent them packing

00:33:52.480 | and they ended up selling to Macromedia

00:33:55.120 | and Macromedia was where Flash was created.

00:33:57.360 | And the good thing about Macromedia was it was a tool company

00:34:02.000 | so it invested in the best ideas, I think,

00:34:05.800 | which are still somewhat lost to us of Flash.

00:34:08.120 | The timeline, animation is sort of an immutable function

00:34:12.500 | over time.

00:34:13.560 | - They had the tooling around that too,

00:34:14.920 | like the Dreamweaver, there's a Flash--

00:34:17.240 | - Flash Director, there were a bunch of them.

00:34:18.840 | - Yeah, I mean, yeah, that was a great--

00:34:20.880 | - Flash Builder was one of the last ones.

00:34:22.800 | These tools were used by real artists

00:34:25.400 | and special effects people and designers.

00:34:27.440 | All the restaurant websites around 2005 were done in Flash

00:34:30.880 | which was, we were trying to do HTML5 at the same time

00:34:34.080 | that was the Firefox era.

00:34:35.080 | We were trying to make the web capable enough

00:34:37.640 | you didn't need Flash but if you recall,

00:34:39.400 | you go to a restaurant and it's like,

00:34:41.080 | this is kind of like a game or something.

00:34:42.400 | It's like a Flash, all the font looks small.

00:34:44.640 | - You didn't like Flash from the beginning.

00:34:45.920 | You're like, this doesn't feel right.

00:34:47.800 | - Not really, I actually admire Flash's technology

00:34:50.040 | and I'm pretty pragmatic about these things

00:34:51.740 | and I realized that it doesn't matter

00:34:54.880 | if you're dealt a bad hand like JavaScript was a rush job

00:34:58.320 | or if you have Flash as a plugin

00:35:00.280 | and you can invest in the tools and make it pretty good.

00:35:03.680 | You should make it better for your users

00:35:05.760 | and grow it as best you can.

00:35:07.040 | And what happened with the browser

00:35:08.760 | due to Microsoft's monopoly abuse

00:35:11.080 | for which they were convicted and even after that

00:35:15.100 | until I think Firefox and then Chrome

00:35:17.480 | was people kept saying, oh, the web can't do X,

00:35:19.960 | it can't do Y.

00:35:21.280 | We'll have to have a plugin.

00:35:22.280 | We'll have to have a new approach.

00:35:23.720 | We'll clean the slate and have a new web

00:35:26.900 | and everyone who said that failed

00:35:29.000 | and the reason they failed is because there's too much value

00:35:31.740 | in the web, this huge network

00:35:34.120 | and the worse is better principle

00:35:35.760 | means that you can not only start bad

00:35:37.900 | which they all sneer at but get on first

00:35:40.880 | and get wide distribution, get sort of evolutionary advantage

00:35:44.560 | and priority of place but you can also improve it over time.

00:35:49.360 | And so if you're gonna improve Flash

00:35:51.120 | and for some reason Flash is now out of favor,

00:35:53.040 | Steve Jobs said you can't have Flash on the iPhone.

00:35:55.040 | That was probably the death knell.

00:35:56.760 | Put your energy into JavaScript and that happened, right?

00:35:58.880 | So we did things at Mozilla with Adobe to improve

00:36:02.440 | which bought my Macromedia to improve Flash

00:36:04.880 | and to improve the version of JavaScript that was in Flash.

00:36:07.560 | We tried to standardize that.

00:36:09.040 | - Oh, that's right.

00:36:09.880 | - I'm getting ahead of myself.

00:36:10.720 | - You can program Flash. - ES4, yeah.

00:36:12.040 | - That's right.

00:36:13.200 | Can we just rewind to the magical,

00:36:15.200 | like it's a special moment in the history

00:36:18.600 | of all of computing.

00:36:20.200 | We'll talk about it later but it's arguable,

00:36:22.880 | it's possible that the entirety of the world

00:36:25.320 | will run in JavaScript at some point.

00:36:27.280 | So it's like those days, it would be interesting

00:36:31.280 | if you could just describe actually zooming in

00:36:34.800 | on how the cake was baked from the several days

00:36:39.800 | that you were working on it, what was on your mind,

00:36:43.360 | how much coffee were you drinking,

00:36:45.400 | were you nervous, freaking out?

00:36:47.360 | - I'll try to remember it.

00:36:48.360 | I mean, you're right, there were these pregnant moments

00:36:50.240 | you see in hindsight, maybe they're overrated

00:36:52.240 | but like Hegel sees Napoleon on horseback at Jena

00:36:55.800 | and says, "There's the world spirit on horse."

00:36:59.240 | And I knew that there was a chance to do it, Mark knew,

00:37:05.080 | and he was my executive sponsor

00:37:08.000 | and he was the one sort of brainstorming

00:37:10.840 | how the JavaScript should be right there in the page.

00:37:13.400 | That was important for him to say that

00:37:14.800 | 'cause I thought so too but a lot of people were like,

00:37:17.320 | "Oh, you can't write programming language

00:37:18.680 | "in the middle of the markup."

00:37:20.120 | And indeed there are problems, if you did it naively,

00:37:22.000 | you'd see the code laid out as like random gibberish.

00:37:25.340 | So I had to figure out how to hide that,

00:37:26.640 | that was a challenge.

00:37:28.160 | - Is that a breakthrough idea?

00:37:29.840 | I mean, so you and Mark thinking about this idea

00:37:32.600 | that you just inject code in the middle of the markup.

00:37:36.160 | - Of the webpage, yeah, it was considered kind of heretical.

00:37:38.040 | There was an SGML guru, I forget his name,

00:37:40.160 | but he corresponded with me and at first he was angry.

00:37:42.120 | He's like, "You should have used a marked section,

00:37:44.680 | "why didn't you use a marked section?"

00:37:45.960 | And I said, "Well, SGML marked sections

00:37:47.680 | "are not part of HTML by the way,

00:37:48.900 | "and they're not supported in the browser."

00:37:50.640 | And so I did some hack that was equivalent

00:37:52.560 | and over time you could do the proper SGML thing.

00:37:55.320 | But eventually he came around and it was again

00:37:58.240 | sort of evolutionary necessity,

00:37:59.680 | it was almost like introgression,

00:38:00.880 | like the idea which Lin Margulies I think helped get across

00:38:07.000 | that we have to consider mutualism biology

00:38:10.640 | that maybe mitochondria were ancient prokaryotes

00:38:14.080 | that got into the cell and became beneficial.

00:38:17.400 | Somehow the same sort of thinking applies,

00:38:21.200 | you have to embed JavaScript in HTML.

00:38:23.040 | It's gonna be a good virus.

00:38:24.480 | - It won't hurt you.

00:38:25.320 | - So the code becomes data in the sense

00:38:27.400 | it just gets carried along.

00:38:30.040 | But is there the side of the,

00:38:32.760 | so you were focusing on the Netscape at that time,

00:38:35.080 | doesn't the browser have to support,

00:38:37.080 | interpret correctly this mix of HTML and whatever code?

00:38:41.440 | - I had to hide it from old browsers including Netscape 1.1

00:38:43.800 | which was predominant then.

00:38:45.140 | So I used an HTML comment,

00:38:48.000 | but inside the container that comment lived in,

00:38:50.360 | the script tag, which is a new element,

00:38:51.980 | I could make different semantics in Netscape 2

00:38:54.640 | where those HTML comment delimiters

00:38:57.040 | instead of being multi-line brackets

00:38:58.520 | became one line or essentially one line.

00:39:00.680 | - So you wrote, so JavaScript was written,

00:39:02.560 | the programming language was written as a comment.

00:39:05.040 | - Yeah, comment for old browsers

00:39:06.120 | and a set of brackets that were ignored

00:39:08.800 | with real code for new.

00:39:10.120 | And it was this two-way comment hiding hack, as I called it,

00:39:13.480 | that was absolutely necessary for us to get off the ground.

00:39:15.720 | We couldn't have bootstrapped JavaScript without it.

00:39:17.680 | We didn't have scripts that were loaded

00:39:19.480 | from a separate file.

00:39:20.320 | The only scripts in Netscape 2 were inline in the document.

00:39:22.920 | - What were the challenges here?

00:39:24.840 | Like what, you know, typing,

00:39:27.420 | what were the choices you were thinking about?

00:39:30.200 | - Garbage collection. - Was garbage collection.

00:39:31.640 | - I didn't have time to write a garbage collector,

00:39:33.000 | so I just, I didn't at first.

00:39:34.680 | So the thing was using essentially arenas

00:39:36.920 | or what GNU calls obj pools

00:39:39.160 | and just would run out of memory eventually.

00:39:41.680 | And I added reference counting in a hurry

00:39:43.560 | after the 10 days in which I hacked.

00:39:45.360 | So after I was in the server team doing HTTP 1.1

00:39:48.120 | and thinking about the language,

00:39:49.640 | I finally got transferred to the client team in early May.

00:39:52.320 | And that's when I, you know, I got the go sign from Mark

00:39:55.280 | and it was like, we can't wait

00:39:57.280 | because people inside Netscape are doubting,

00:39:59.360 | even people inside Sun are definitely doubting.

00:40:01.880 | Bill Joy was the champion, but he was like alone in that

00:40:04.960 | in seeing there was a role for JavaScript

00:40:07.240 | as the, as I call it, the sidekick language,

00:40:10.560 | robbing the boy hostage.

00:40:11.960 | (laughing)

00:40:13.080 | Frank Miller put it in the Dark Knight Returns

00:40:16.360 | that there was this silly little language

00:40:18.640 | that would be the glue language

00:40:19.480 | and it could become important over time.

00:40:20.960 | And you were better off having that complementarity,

00:40:23.880 | that pairing of languages,

00:40:25.120 | just like Microsoft stacked it

00:40:26.960 | with Visual C++ and Visual Basic.

00:40:28.880 | - So what was the big moment of I'm done?

00:40:33.120 | - So I had to do a demo.

00:40:34.320 | I forget the dates.

00:40:35.480 | I think I, for a history of programming languages paper

00:40:38.680 | that Alan Weissbrock did with my help,

00:40:40.840 | he did a lot of the writing.

00:40:42.720 | I think it was the 10 days from like Thursday evening

00:40:47.440 | through to the following weeks, you know,

00:40:50.800 | the whole of that week and then into the Monday.

00:40:52.440 | - Did you get sleep?

00:40:53.280 | - Not, not enough.

00:40:54.400 | And I was really going fast

00:40:56.280 | because I'd already used a lot of C compiler

00:41:00.080 | and front-end compiler knowledge

00:41:02.320 | that I'd gained from undergraduate school.

00:41:04.680 | When I started getting into computing

00:41:06.560 | as a renegade physics major,

00:41:10.040 | people were formalizing more efficient bottom-up grammars,

00:41:14.640 | parsers for bottom-up languages.

00:41:17.040 | Really, LALR1 was the big thing.

00:41:19.520 | And I studied all this and learned how to parse them.

00:41:22.440 | And in the end, if you're doing C languages,

00:41:25.100 | you often do what Dennis Ritchie did anyway,

00:41:28.640 | which is a recursive descent parser.

00:41:31.280 | You can hand code it.

00:41:32.200 | And I did that for JavaScript in a blazing hurry.

00:41:36.280 | Mostly got it right.

00:41:37.360 | Didn't have precedence inversion problems or other bugs,

00:41:40.160 | but I copied a lot from Java and C.

00:41:43.160 | And I tried to keep things simple.

00:41:45.240 | Like the equality operator in those 10 days sprint

00:41:49.400 | between two objects of different dynamic type said,

00:41:52.280 | "No, they're not equal.

00:41:53.120 | Their types are different."

00:41:54.320 | And then after that, I had internal early adopters

00:41:57.660 | and they were using JavaScript to match a number

00:42:01.260 | against a database field that had been stringized.

00:42:03.920 | And they said, "Oh, can't we just have implicit conversion?"

00:42:05.920 | And like an idiot, I agreed.

00:42:07.700 | I gave them what they wanted.

00:42:08.620 | I was trying to please them and get adoption.

00:42:10.640 | And that broke what equivalence relation nature

00:42:15.640 | there was to the double equal.

00:42:19.600 | There's some edge cases with not a number

00:42:22.360 | that break that at two.

00:42:23.320 | But it really broke it.

00:42:25.040 | Having implicit conversions in the operator

00:42:26.960 | is something that people still roast me over.

00:42:29.320 | - So let's talk about two things.

00:42:31.720 | One, it sounds like the comparison operator,

00:42:34.280 | the equality operator is the thing that you regret.

00:42:38.760 | So maybe can you--

00:42:39.600 | - Making it sloppy.

00:42:40.600 | - Making it sloppy.

00:42:41.440 | So what is the biggest thing you regret in those 10 days?

00:42:45.200 | And what is the biggest thing you're proud of?

00:42:47.480 | - So that making it sloppy came after the 10 days.

00:42:49.680 | And my lesson there, which I've tweeted is,

00:42:51.880 | when people come to you saying,

00:42:53.080 | "Can you please make it sloppy or add this cute feature?"

00:42:56.080 | The answer should be no.

00:42:57.640 | And I should have known that 'cause I think Niklaus Wirth,

00:43:00.120 | one of my heroes, said,

00:43:01.120 | "The essence of design is leaving things out."

00:43:03.640 | But during the 10 days, I also, like I said,

00:43:06.180 | I was in such a hurry, I left out garbage collection.

00:43:08.560 | Came back to haunt me,

00:43:09.400 | but I got reference counting in in time

00:43:11.240 | that people weren't running out of memory right away

00:43:14.400 | on long-lived JavaScript.

00:43:16.080 | - Wait, what happens when you don't have

00:43:17.000 | garbage collection and you have objects?

00:43:18.520 | - Well, you just run out of memory.

00:43:20.160 | (Lex laughs)

00:43:21.000 | And you know, at first, you write a short script

00:43:24.120 | and the page doesn't last long or it doesn't do a lot,

00:43:25.920 | and it's okay.

00:43:26.740 | - Oh, I see, yeah, yeah, yeah.

00:43:27.580 | - But if you're writing a game or something

00:43:28.600 | and you're doing event-based allocation,

00:43:30.600 | you run out of memory.

00:43:31.640 | And this was noticed in the summer of 1995,

00:43:34.320 | and people were like, "What's going on?"

00:43:35.560 | I was, "Oh yeah, I better go back and do reference counting."

00:43:38.600 | And then the problem with reference counting is

00:43:40.200 | you're writing the language in the runtime in C,

00:43:43.160 | an unsafe language, and if you're reference counting

00:43:45.760 | and you overflow the counter, you mismanage it

00:43:48.060 | so it goes high, it gets stuck high,

00:43:49.600 | you leak memory again and you run out.

00:43:51.560 | If you underflow it, you pre-memory that's still in use.

00:43:56.200 | And even then, we knew what all the security hackers

00:44:00.120 | came to know, that you therefore have

00:44:02.600 | potentially a remote code execution vulnerability.

00:44:05.840 | 'Cause this was before things like

00:44:08.120 | non-executable heap memory and

00:44:11.120 | stack defenses against taking over memory.

00:44:14.800 | So if you can, from the remote side,

00:44:17.780 | write some HTML and JavaScript that just happens

00:44:19.900 | to exploit a bug in memory safety,

00:44:21.820 | like it causes JavaScript to underflow a reference counter,

00:44:24.960 | and the script still has its hands on that object

00:44:27.560 | and it's trying to call a method on it,

00:44:29.040 | and there's some kind of lookup function table

00:44:30.680 | in the object, but you've managed to stuff the heap

00:44:34.620 | with strings that forwards their own lookalike

00:44:37.160 | for the function table, you can call some other code.

00:44:41.360 | And this was a problem right away.

00:44:43.400 | So security, JavaScript upped the ante.

00:44:46.280 | Java had this problem too, but in its own VM.

00:44:48.840 | And it just was a separate headache for Sun to worry about.

00:44:53.440 | We had this problem in Netscape right away.

00:44:55.280 | So Netscape 2 came out after my 10 days,

00:44:58.340 | and after these follow-on work to embed JavaScript

00:45:02.640 | better in the browser and to add garbage collection

00:45:05.120 | through reference counting, really,

00:45:06.320 | I call it reference counting, and get it shipped.

00:45:09.660 | We had a bunch of dot releases where we fixed

00:45:11.400 | security bugs like maniacs. (laughs)

00:45:13.560 | - But what is the thing you're,

00:45:16.000 | you know, when you sit back on a porch

00:45:18.360 | and just look out into the sunset,

00:45:20.000 | what are you most proud of from those 10 days?

00:45:21.960 | - I think the first class functions, Sean.

00:45:23.560 | I think, especially since Java didn't have it

00:45:25.640 | and it was somewhat unusual.

00:45:27.600 | - Scheme made it in somehow at the end of the day.

00:45:30.640 | - In spirit, in spirit.

00:45:31.480 | I mean, people complain because Scheme has,

00:45:33.440 | you know, minimalism.

00:45:34.840 | It has six or seven special forms.

00:45:37.000 | It has hygienic macros.

00:45:38.720 | It has call CC.

00:45:39.680 | It has sort of a beautiful, complete set of forms

00:45:44.680 | to make the Lambda calculus pleasant to use in practice.

00:45:48.640 | And JavaScript is, you know,

00:45:51.640 | kind of a multi-paradigm or Schembolic. (laughs)

00:45:56.120 | - Just on a small tangent, you mentioned Mark Andreessen.

00:45:59.400 | It sounds like, and Bill Joy, but staying on Mark,

00:46:03.400 | it sounds like he had an impact on you

00:46:06.520 | in that he sort of believed in what you were doing there.

00:46:09.120 | Can you talk about like what role Mark had in your life?

00:46:11.920 | - Yeah, we would meet at the Pinsla Creamery

00:46:15.880 | in downtown Palo Alto.

00:46:17.720 | And Mark was just fresh out of, you know, grad school

00:46:19.960 | or whatever he was doing.

00:46:20.880 | And he was a big dude and he got fitter later.

00:46:25.280 | He had hair.

00:46:26.120 | He would order giant milkshakes and burgers.

00:46:28.800 | And we would meet there and brainstorm about what to do.

00:46:31.040 | And it was very direct because we didn't have much time.

00:46:34.020 | Sort of, we didn't talk about it,

00:46:36.280 | but the implication was Microsoft was coming after us.

00:46:39.080 | Mark was saying things boldly, pre-IPO,

00:46:41.760 | like Netscape plus Java kills Windows, right?

00:46:44.320 | - Wow, it's ambitious.

00:46:45.640 | - Make a browser programmable,

00:46:46.840 | it becomes the new runtime for programs.

00:46:49.400 | It's the meta OS or it's the replacement OS.

00:46:52.000 | - But he still saw value in JavaScript.

00:46:54.280 | - Yes, even though he was saying that

00:46:56.000 | and Java was the big name, hence the trademark license,

00:46:59.800 | he saw JavaScript as important.

00:47:01.360 | And he even thought, what if we got,

00:47:03.720 | I told this in other interviews, I can say it,

00:47:05.040 | he thought, what if we had my friend Kip Hickman,

00:47:07.660 | who'd been at Netscape from the beginning

00:47:09.520 | and who was a kernel hacker at SGI when I joined,

00:47:13.080 | he started writing his own JVM

00:47:14.880 | before we consummated the Sun deal

00:47:16.400 | and got our hands on their code.

00:47:17.960 | And the Java compiler, Java C,

00:47:20.480 | which Arthur Van Hoff had written, very nice code,

00:47:23.200 | was all written in Java, it was self-hosted

00:47:24.960 | or so-called bootstrap.

00:47:26.480 | And so we could use that as soon as Kip's Java VM

00:47:29.320 | could run the bytecode from the Sun JVM

00:47:34.320 | running the self-hosted compiler to emit the bytecode.

00:47:37.520 | So once we could bootstrap into Kip's VM,

00:47:39.420 | we wouldn't need Sun.

00:47:40.300 | And Mark was like, well, maybe we can just ditch Sun,

00:47:44.380 | we'll have Kip's Java VM, we'll have your JavaScript VM.

00:47:47.340 | Now we need graphics.

00:47:48.220 | So Mark was thinking far ahead

00:47:49.600 | because he knew you could do things with HTML and images,

00:47:53.380 | but at some point you really want--

00:47:54.900 | - Like dynamo graphics or three-dimensional?

00:47:57.340 | - Like even SGI had already started its downfall

00:47:59.920 | because the first four VLSI team there

00:48:02.940 | had gone off to do 3DFX and all these other companies

00:48:05.380 | that made the graphics card on your PC, right?

00:48:07.480 | Doom was big and Quake.

00:48:09.360 | And so we were all playing Quake.

00:48:10.660 | I was old, so I was terrible.

00:48:12.860 | But why not put that graphics capability on the web?

00:48:16.360 | And in fact, it finally happened at Mozilla

00:48:18.680 | with Firefox era with Vlad Vukicevic

00:48:21.480 | taking OpenGL ES and reflecting it as WebGL.

00:48:24.380 | But OpenGL ES is the mobile version of OpenGL,

00:48:27.960 | which is a standard based on SGI GL.

00:48:30.300 | So this whole lineage of graphics libraries

00:48:32.760 | or really graphics languages for what became the GPU.

00:48:36.260 | And Mark was thinking ahead.

00:48:37.300 | He's like, "We need graphics too."

00:48:38.540 | And I thought, "Okay, I can try to get somebody

00:48:40.300 | I knew at SGI, but he's a grad student at MIT.

00:48:42.860 | He was studying under Barbara Liskov.

00:48:44.500 | He laughed when he heard about this later.

00:48:46.140 | Andrew Myers, he's at Cornell, long time.

00:48:48.420 | I think he's a full professor.

00:48:50.100 | And Mark said, "Great, we'll get him."

00:48:51.740 | And I'm like, "I'm not sure he's gonna come.

00:48:53.140 | We'll throw money, we'll stock options."

00:48:55.260 | We never did it.

00:48:56.120 | And they did the Sun deal.

00:48:57.580 | So Kip Nobly put aside his own JVM

00:49:00.900 | and we used the Sun JVM.

00:49:02.860 | So that was an ambitious period.

00:49:03.980 | And Mark was very generative because he was pushing hard.

00:49:06.620 | He was ambitious and he wanted to have Netscape

00:49:11.180 | possibly be in control of the ball.

00:49:13.660 | - Maybe you can speak to this dance

00:49:16.860 | of Netscape versus Internet Explorer.

00:49:20.500 | You've thrown some loving words towards Microsoft

00:49:25.500 | throughout this conversation,

00:49:26.700 | but that's a theme with, I mean,

00:49:28.440 | Steve Jobs had a similar sort of commentary.

00:49:31.260 | From a big sort of philosophical principle perspective,

00:49:34.500 | can you comment on the approach that Microsoft has taken

00:49:37.780 | with Internet Explorer from IE1 to Edge today?

00:49:42.100 | Is there something that you see as valuable

00:49:45.380 | that they're doing in the occasional copying

00:49:48.380 | and that kind of stuff?

00:49:49.580 | Or is the world worse off because Internet Explorer exists?

00:49:54.580 | - So I'm gonna segment this into historical eras

00:49:57.620 | because I think Microsofters today with Satya

00:49:59.560 | is quite a different company

00:50:00.620 | and what they're doing with Edge is different.

00:50:02.340 | But back then, Gates, aggressive character,

00:50:06.160 | not really original in my view, not an originator.

00:50:09.660 | Steve Jobs famously said once, "He doesn't have any taste."

00:50:12.420 | And I don't mean this in a small way.

00:50:13.780 | He has no taste.

00:50:15.060 | You can see this.

00:50:16.540 | Apple at the time had beautiful typography

00:50:19.340 | and ligatures and kerning and the fonts looked great.

00:50:22.580 | And Windows had this sort of ugly system font

00:50:25.220 | that was carefully aligned with pixels

00:50:26.940 | so it didn't get anti-aliased.

00:50:28.660 | - What is it?

00:50:29.500 | I'm sorry to keep interrupting.

00:50:30.320 | Why was Internet Explorer winning

00:50:32.380 | throughout the history of these competitions?

00:50:34.300 | - Distribution.

00:50:35.220 | Distribution matters more than anything.

00:50:36.620 | And this is why even now we're seeing

00:50:39.820 | in the browser wars, Edge doing better

00:50:41.660 | because it's being foisted on people of Windows.

00:50:43.460 | We have Windows 10 boxes at home.

00:50:45.220 | We have some Windows 7 boxes or laptops we keep running too

00:50:48.220 | because we don't connect them to the internet generally.

00:50:50.860 | But once you have that operating system to hold,

00:50:56.140 | you can force Edge.

00:50:57.820 | And Apple did it with Safari too.

00:50:59.500 | It's not unique to Microsoft.

00:51:01.340 | - That's sad.

00:51:02.180 | - But distribution matters.

00:51:03.660 | And that's why I think IE was going to win.

00:51:07.060 | That's why everybody at Netscape felt we're doomed.

00:51:09.580 | This was something Michael Toy and Jamie Woodson were doomed.

00:51:12.900 | But for a while there, we had a chance

00:51:14.700 | and we innovated in Netscape too.

00:51:16.060 | We did a big platform push.

00:51:17.540 | Java and JavaScript and plugins, more plugins,

00:51:21.740 | and more HTML table features.

00:51:25.140 | And really started making a pre-arable stack

00:51:28.700 | out of what were pretty static web languages.

00:51:30.820 | And even in the beta releases of Netscape,

00:51:33.220 | two people were using JavaScript

00:51:34.460 | to build what you would call single page applications

00:51:36.620 | like Gmail.

00:51:37.740 | And they were using JavaScript locally to compute things

00:51:40.380 | and to call the server on a hidden frame in the background.

00:51:43.660 | So it was prefiguring a lot of what came later as AJAX

00:51:46.860 | or dynamic JavaScript, dynamic HTML.

00:51:48.620 | - So people saw that, I mean--

00:51:49.740 | - Even then they saw it, yeah.

00:51:50.900 | - That's kind of, I don't know, from my perspective,

00:51:53.820 | that seems quite brilliant.

00:51:54.940 | It seems like really innovative

00:51:56.580 | that you would have code run in the browser.

00:51:59.220 | - It did impress me with something,

00:52:01.380 | which I learned later about from Eric von Hippel of MIT,

00:52:04.020 | which is user innovation networks, lead user effects.

00:52:07.260 | That throwing out JavaScript,

00:52:09.180 | even though we weren't doing open source,

00:52:10.460 | we were doing beta releases early

00:52:12.300 | and permissively with Netscape.

00:52:14.340 | Getting early developer feedback, absolutely critical.

00:52:17.180 | I loved it.

00:52:18.020 | I did some of that with SGI,

00:52:19.140 | with some of the products I worked on,

00:52:20.260 | but it really came to the fore in Netscape.

00:52:21.860 | And that culminated in Mozilla,

00:52:23.740 | where you're dealing with developers all the time

00:52:25.700 | and early adopters, lead users.

00:52:27.200 | But the lead users helped improve JavaScript,

00:52:29.800 | even in those last few betas,

00:52:31.540 | where I could hardly change things.

00:52:32.620 | I was under pretty rigid change control.

00:52:34.540 | - So we're talking about just a small collection

00:52:36.220 | of individuals that are just like up front--

00:52:38.260 | - A guy named Bill Dortch,

00:52:39.380 | you can find his work in the web archive,

00:52:40.900 | still from 1996.

00:52:41.940 | It's a single page application.

00:52:43.240 | It's an artist gallery of mountain art.

00:52:46.300 | - He used JavaScript.

00:52:47.140 | - It doesn't quite work.

00:52:47.960 | He uses JavaScript locally.

00:52:48.800 | He uses a local database.

00:52:50.540 | What you would think of now is JSON,

00:52:52.180 | but it's all pure JavaScript code,

00:52:54.260 | a bunch of objects being constructed.

00:52:55.820 | - That's so cool.

00:52:57.460 | So how is, if you can do sort of a big,

00:53:00.780 | sweeping progress of JavaScript,

00:53:03.220 | how has JavaScript changed over the years,

00:53:05.380 | in your view, from those early 10 days,

00:53:07.320 | with a quick addition of garbage collection

00:53:09.460 | and fixes around security?

00:53:11.140 | How has this evolution, that now it's taken over the world?

00:53:14.140 | - It's been a bumpy ride,

00:53:16.340 | because the standards body got shut down after Microsoft,

00:53:19.900 | I think, took over the web,

00:53:21.540 | and then felt punished by the US v. Microsoft

00:53:24.220 | antitrust case.

00:53:25.380 | - Can you speak to the standard body?

00:53:27.380 | - That was a fun ride, too,

00:53:28.260 | because Netscape had taken the lead with the web

00:53:32.300 | and HTML innovations like frames and framesets tables.

00:53:37.100 | And the W3C was sort of off,

00:53:38.740 | even then, sort of in SGML land,

00:53:40.460 | heading toward XML la-la land.

00:53:43.100 | I'm gonna be a little harsh on it.

00:53:44.340 | - What's SGML?

00:53:45.180 | I'm sorry.

00:53:46.000 | - SGML was the precursor markup language to HTML,

00:53:48.900 | or it was sort of the more extensible

00:53:51.020 | standard generalized markup language.

00:53:54.140 | It was a-- - XML-like.

00:53:55.820 | - Pointy brackets, but it had all sorts of elaborate syntax

00:53:58.540 | for doing different semantics.

00:54:00.940 | And this is why I think TBL and others

00:54:03.780 | who wanted to do the semantic web

00:54:05.580 | then took XML forward.

00:54:08.340 | But they had this, or some of them anyway,

00:54:10.420 | had this strange idea that they could replace the web

00:54:12.340 | with XML, or that they would upgrade the web to be XML.

00:54:15.740 | And it couldn't be done.

00:54:17.260 | Worse is better had concrete meaning.

00:54:19.540 | The web was very forgiving of HTML,

00:54:22.660 | including sort of minor syntax errors

00:54:25.220 | that could be error corrected.

00:54:26.500 | Like error correction isn't generally done

00:54:28.180 | in programming languages because--

00:54:29.380 | - Right.

00:54:30.220 | That's another amazing thing about HTML,

00:54:31.860 | is like it's more like biology than programming.

00:54:34.820 | - Yeah, exactly.

00:54:36.100 | And so XML was, in its standard form,

00:54:40.060 | super strict and could never have admitted

00:54:42.580 | the kind of users who were committing these errors.

00:54:44.460 | And the funniest part was Microsoft said,

00:54:46.540 | "Hey, we're doing XML."

00:54:47.660 | But the way they put it in Internet Explorer

00:54:49.580 | under the default media type,

00:54:51.540 | put it through the HTML error corrector.

00:54:53.220 | (laughing)

00:54:54.060 | - Oh, wow.

00:54:55.020 | - So they kind of bastardized it

00:54:56.340 | to make it popular and usable and accessible.

00:54:58.620 | And so XML as a pure thing was never gonna take over.

00:55:03.620 | And W3C was kind of not fully functional

00:55:07.220 | because Netscape wasn't cooperating with them.

00:55:09.780 | We thought about where to take JavaScript

00:55:11.180 | and we realized, our standards guru, Carl Cargill,

00:55:13.980 | realized there was a European standards body

00:55:17.140 | that had already given Microsoft fits

00:55:18.820 | by standardizing parts of the Windows 3.1 API,

00:55:22.580 | which European governments insisted on.

00:55:24.020 | They said, "Microsoft, we can't use your operating system

00:55:26.180 | "without some standards."

00:55:27.620 | And Microsoft said, "Here's our docs."

00:55:29.660 | And the government said, "No, we need a European standard."

00:55:32.460 | So this body called

00:55:33.420 | the European Computer Manufacturers Association, ECMA,

00:55:37.500 | which eventually became global

00:55:38.580 | and became a proper noun instead of an acronym.

00:55:40.980 | - Right, it's just one capital E now

00:55:43.260 | with a lowercase CMA.

00:55:44.660 | - Right.

00:55:45.500 | And as one of the early Microsoft guys I met

00:55:47.580 | when we first convened a working group

00:55:50.260 | to talk about JavaScript said,

00:55:51.340 | "It sounds like a skin disease."

00:55:53.420 | - But it gave, I mean, maybe you'll speak to that,

00:55:56.140 | but it gave the name to JavaScript of ECMAScript.

00:55:58.980 | - That was the standard name

00:55:59.820 | because Java was a trademark of Suns.

00:56:02.820 | They were so aggressive,

00:56:03.700 | they were sending cease and desist letters

00:56:05.820 | to people whose middle European heritage

00:56:08.500 | meant their surname was Javanko

00:56:10.780 | and they called their website javanko.com

00:56:12.860 | and Sun would send them a letter saying,

00:56:14.580 | "You're using J-A-V-A at the start of your domain name.

00:56:16.700 | "You must cease and desist."

00:56:18.900 | - I love marketing more than anything else in this world.

00:56:22.020 | So ECMAScript and now is popularly named as ES plus version.

00:56:27.020 | - I would say people use JS more than anything.

00:56:29.860 | People still say JavaScript.

00:56:31.100 | JavaScript is in all the books.

00:56:32.380 | - So I mean, when you're referring to it,

00:56:33.700 | it's usually JavaScript

00:56:34.860 | and when you wanna refer to a version of JavaScript,

00:56:37.300 | you'll say ES6, ES5.

00:56:39.140 | - Yes, or now they've gone to years,

00:56:40.820 | which is kind of confusing

00:56:41.820 | 'cause it's an offset of 2009.

00:56:45.500 | ES6 is ES2016.

00:56:49.220 | - Yeah, it doesn't match the years perfectly.

00:56:52.700 | Yeah, so what were the choices made

00:56:55.660 | and how did JavaScript evolve here?

00:56:57.980 | - So we took this new standards body,

00:57:01.060 | which we thought sort of a proven record

00:57:04.340 | of standing up to Microsoft,

00:57:05.380 | but Microsoft sent a lot of people.

00:57:07.540 | They sent some people who were pretty good

00:57:10.540 | and then when they realized that I was there

00:57:12.940 | and Netscape was not gonna just bend over

00:57:15.220 | and do whatever they wanted,

00:57:16.380 | they sent somebody really good.

00:57:17.300 | He was a smart guy.

00:57:18.140 | He did a lot of the work on the first draft of the spec,

00:57:21.020 | Sean Katzenberger, he's left Microsoft.

00:57:22.780 | He even did what I sort of did.

00:57:25.260 | He told his bosses, stop bugging me to do other things.

00:57:27.980 | I'm focused on this

00:57:28.940 | 'cause it took a lot of focused work

00:57:30.580 | to create the first draft of the spec.

00:57:31.940 | And I was still holding,

00:57:33.540 | I was spending almost all the plates.

00:57:34.900 | I had like part-time help in certain areas

00:57:37.220 | and on the front end integrations,

00:57:38.740 | I had the front end guys.

00:57:40.100 | But I couldn't take as much time as Sean was

00:57:42.620 | to write the draft spec,

00:57:43.660 | but I had to participate

00:57:44.660 | because I was essentially helping write down

00:57:46.880 | what the language did.

00:57:48.020 | And in areas where we didn't like what it did

00:57:50.180 | and Microsoft didn't agree,

00:57:51.620 | we sometimes got away with slight changes.

00:57:54.580 | And that's the story of standards.

00:57:55.780 | You have different implementations

00:57:57.980 | and depending on their market power,

00:57:59.820 | they interoperate where you have agreement

00:58:01.820 | and where they don't,

00:58:02.780 | the dominant one usually sets the de facto standard.

00:58:05.900 | And then you should probably reflect that

00:58:07.500 | into the de jure standard.

00:58:09.340 | And this happened with JavaScript.

00:58:11.060 | Over time as Netscape went down and Microsoft went up,

00:58:14.140 | we did the first edition of the standard codified in 1997.

00:58:18.180 | In France, we had a trip to Nice,

00:58:20.540 | which was very memorable.

00:58:22.140 | - For any interesting reason or just because it's Nice?

00:58:24.540 | - And ECMA's European and IBM and others were there.

00:58:28.980 | Mike Kalashaw, an IBM fellow was a British.

00:58:31.340 | And the guy who ran ECMA at the time,

00:58:34.460 | Jan van den Bel was quite a raconteur

00:58:37.740 | and a very fun guy.

00:58:39.060 | And he had us out for the great,

00:58:41.540 | "Fui de mer," the bouillabaisse.

00:58:43.500 | - Was the standardization process beautiful or painful

00:58:46.140 | that those early days, you as a designer of the language?

00:58:48.500 | - It was painful because it was rushed.

00:58:50.140 | Now, Guy Steele was contributed by Sun.

00:58:52.420 | So even more than Sean,

00:58:53.620 | you had this giant brain, Guy Steele helping,

00:58:55.980 | bringing some of that scheme magic.

00:58:57.220 | He even brought Richard Gabriel for fun.

00:58:59.500 | Richard wrote the fourth clause of the ECMA standard,

00:59:02.900 | which was kind of an intro to what JavaScript's all about.

00:59:05.260 | So we had some really good people.

00:59:07.220 | And we didn't fight too much.

00:59:09.100 | There was some tension where I was fixing bugs

00:59:11.460 | and I was late to a meeting.

00:59:12.700 | And Sean Katzenberger of Microsoft was actually mad.

00:59:14.980 | Like, "Where is he? We need him."

00:59:16.580 | And when I got there,

00:59:18.620 | I saw that only he saw this sort of off by one bug

00:59:21.860 | and somewhere in the spec.

00:59:22.920 | And then I saw it too.

00:59:23.980 | And I said, "There's a fence post bug there."

00:59:25.260 | And then we kind of locked eyes

00:59:26.980 | and we realized we were on the same page.

00:59:28.460 | And we kind of, he wasn't mad anymore.

00:59:30.260 | - What were the features that are being struggled over

00:59:33.020 | and debated and thought about?

00:59:34.980 | - It was mainly writing down what worked

00:59:36.500 | and what we thought should work in the edge cases

00:59:38.420 | that didn't interoperate or that seemed wrong.

00:59:41.540 | But we were already laying the groundwork

00:59:43.100 | for the future editions that I was already implementing.

00:59:46.220 | I was still trying to lead the standard

00:59:48.100 | by using the dominant market power

00:59:50.100 | to write the code that actually shipped.

00:59:52.500 | So the de facto standard would lead the de jure standard.

00:59:54.740 | And I was putting in the missing function forms

00:59:58.220 | that I didn't have time for in the 10 days.

01:00:00.500 | - So this is the engineering mindset

01:00:02.180 | versus the theoretician.

01:00:03.500 | So you didn't want to create the perfect language,

01:00:05.120 | but one that was popular and shipped

01:00:06.940 | and all that kind of stuff.

01:00:07.780 | - And you could say there was,

01:00:08.620 | I was standing on the shoulders of giants.

01:00:09.940 | So there was a staged process where I had to hold back

01:00:12.500 | things that were well designed by others

01:00:15.220 | in other languages that I could imitate.

01:00:17.260 | But I couldn't do them all in the 10 days.

01:00:18.540 | So they came in in 1996 and '97.

01:00:22.820 | And they came into the third edition of the standard,

01:00:25.140 | which was finalized in 1999.

01:00:27.680 | But at that point, Netscape had been sold to AOL

01:00:30.460 | and was, which was a decent exit considering.

01:00:33.140 | And had previously been mercilessly crushed.

01:00:37.400 | Netscape was selling the browser along with server software

01:00:40.680 | that it had acquired after its IPO.

01:00:42.760 | And Microsoft was just underpricing it.

01:00:44.560 | So there was no way to compete with that.

01:00:47.480 | Microsoft was also making Internet Explorer

01:00:51.440 | the default browser in Windows,

01:00:53.140 | which is called tying in antitrust law.

01:00:55.560 | And they were doing even more brutal things.

01:00:57.140 | There's a famous investor, he did very well on Google.

01:01:00.340 | So he's a billionaire, Ram Shriram.

01:01:01.800 | He was sales guy or head of sales at Netscape.

01:01:04.680 | And he got off the phone looking ashen faced

01:01:07.920 | after Compaq called and said,

01:01:09.620 | "Microsoft just told us they're gonna pull

01:01:12.000 | "our Windows license if we ship Netscape

01:01:13.800 | "as the default browser."

01:01:15.080 | - Wow.

01:01:15.920 | So there is some bullying going on.

01:01:17.440 | - It was totally immaterial in the antitrust case.

01:01:19.720 | But JavaScript escaped into the standard setting

01:01:23.080 | where there was fairly good cooperation.

01:01:24.900 | Microsoft had a really good guy on it.

01:01:26.880 | And Guy Steele was there for a time.

01:01:29.440 | And there was some good work.

01:01:30.920 | But after the antitrust case and Netscape

01:01:35.200 | kind of dissolving into AOL

01:01:37.480 | and not really going anywhere quickly,

01:01:38.920 | Mozilla took years to really bring up,

01:01:40.940 | the standard froze.

01:01:43.400 | And by 2003, even though they'd been sort of noodling around

01:01:46.300 | with advanced versions, JavaScript 2,

01:01:48.880 | I'd given the keys to the kingdom

01:01:50.280 | to another MIT grad, Baltimore Horwatt.

01:01:53.040 | Very big brain, still at Google, I think.

01:01:54.880 | He won the Putnam in '86.

01:01:56.600 | So he's, yeah, very mathematical.

01:01:58.360 | - Legit.

01:01:59.600 | - He designed the successor language, JavaScript 2,

01:02:02.960 | but it only showed up in mutated form

01:02:06.600 | in Microsoft's ASP.NET server side.

01:02:09.560 | And it didn't last there.

01:02:11.120 | And it showed up in Flash.

01:02:12.320 | And that's what became ActionScript 3.

01:02:14.640 | - Ah, ActionScript.

01:02:15.960 | Interesting.

01:02:16.800 | And then Flash was declined.

01:02:18.720 | And so how did we arrive at ES6,

01:02:21.280 | where it's like, there's so many,

01:02:24.080 | where everyone, okay, there's this history of JavaScript

01:02:27.200 | that people were, it was just like, cool,

01:02:29.200 | when you're having beers to talk crap about JavaScript.

01:02:32.520 | Everyone loves to hate, like people who are married say,

01:02:35.280 | ah, marriage sucks, is they just wanna let off some steam,

01:02:38.880 | even though everyone uses the language.

01:02:40.880 | But ES6, it's become this,

01:02:43.480 | it fixed major pain points, I think.

01:02:49.200 | - It added things to the language

01:02:50.400 | and added something that was already ES5 strict mode,

01:02:53.600 | but made it implicit in class bodies and module bodies.

01:02:57.280 | It was a big jump, but it accumulated

01:02:59.480 | some of the ES4 designs that we'd done with Adobe

01:03:03.640 | for what we hoped would be the fourth edition of ECMAScript

01:03:07.680 | that were supposed to fold in some of these old JavaScript,

01:03:10.520 | two ideas that had come into ActionScript 3.

01:03:13.080 | So you look at the family tree and you see these forks,

01:03:15.240 | and the main ones are the ones that go into Adobe Flash,

01:03:20.240 | acquired from Acromedia,

01:03:22.040 | and the one that went into the server side

01:03:23.560 | of Microsoft's stack, which kind of died.

01:03:27.060 | And then trying to bring them back into the standard,

01:03:28.840 | and not quite succeeding, ES4 was mothballed.

01:03:32.240 | But all the good parts that everyone liked

01:03:34.300 | made it into ES6.

01:03:35.800 | And so that was a success.

01:03:37.000 | And I said earlier, I had the wrong year,

01:03:38.320 | I think it's 2015, so it's off by--

01:03:40.360 | - For ES6.

01:03:41.200 | - Yeah, it was done, finalized in 2015.

01:03:44.280 | It took a little longer than we hoped,

01:03:45.560 | but 'cause ES5 was 2009,

01:03:48.560 | and that was a smaller increment from ES3.

01:03:51.540 | We skipped four again, we mothballed it.

01:03:53.520 | And we had a split in the committee

01:03:55.480 | where some people said, ES4 is too big,

01:03:57.480 | we're gonna work on incremental improvements,

01:03:59.540 | no new syntax in particular, they promised.

01:04:02.580 | Not quite true, but they added a bunch of interesting APIs,

01:04:06.300 | Alan Weir-Sprock, my co-author of the Hubble paper.

01:04:09.880 | And he was at Microsoft at the time,

01:04:11.140 | I ended up hiring him at Mozilla,

01:04:12.300 | he wanted to get to Mozilla and keep doing

01:04:15.260 | the sort of editor job of the JavaScript standard,

01:04:19.660 | ECMAScript.

01:04:21.300 | And when we got ES6 done, it was a little late,

01:04:24.640 | 2015, and we switched to year numbers.

01:04:27.180 | So people still call it ES6, I call it ES6.

01:04:30.100 | But if you remember, off by nine, plus 2000.

01:04:33.100 | - Yeah, I mean, ES6 is such a big job.

01:04:35.780 | I mean, like you said, there's a thread

01:04:37.620 | that connects all of it, but ES6 is one

01:04:39.260 | that's like, became this language

01:04:41.100 | that almost feels ready to take over the world completely.

01:04:45.180 | - More programming in the large features,

01:04:46.660 | more features you need for larger teams.

01:04:49.380 | - Software engineering.

01:04:50.300 | - Microsoft did something smart too,

01:04:52.240 | Anders and company, Luke Hoban, who's left Microsoft,

01:04:56.240 | also did TypeScript.

01:04:57.800 | And they realized something, I think,

01:05:00.400 | that Gilad Brak has also popularized,

01:05:02.760 | and he was involved in Dart at Google.

01:05:05.720 | If you, don't worry about soundness in the type system,

01:05:08.840 | you don't try to enforce type checks

01:05:10.920 | at runtime in particular, just use it

01:05:12.280 | as sort of a warning system, a tool time type system,

01:05:15.040 | you can still have a lot of value for developers,

01:05:17.680 | especially in large projects.

01:05:19.280 | So TypeScript's been a roaring success for Microsoft.

01:05:21.240 | - What do you think about TypeScript?

01:05:24.120 | Is it adding confusion or is it ultimately beneficial?

01:05:27.360 | - I think it's beneficial.

01:05:28.200 | Now, it's technically a superset of JavaScript,

01:05:30.920 | so of course I love it.

01:05:32.080 | (laughs)

01:05:33.360 | The shortest JavaScript program

01:05:35.560 | is still a TypeScript program.

01:05:36.640 | Any JavaScript program is a TypeScript program,

01:05:38.520 | which is brilliant 'cause then you can start

01:05:40.020 | incrementally adding type annotations,

01:05:41.800 | getting warnings, learning how to use them.

01:05:44.440 | Microsoft's had to kind of look around corners

01:05:47.160 | at the standards body and guess how their version

01:05:49.920 | of modules or decorators should work.

01:05:52.480 | And the standards body then may change things a bit.

01:05:55.400 | So I think they're obligated with TypeScript

01:05:57.640 | either to carry their own version

01:05:59.320 | or to bring it back with incompatible changes

01:06:01.680 | towards the standard over time.

01:06:03.120 | And I think they've played generally fair there.

01:06:05.440 | There's some sentiment that,

01:06:06.480 | why don't they standardize TypeScript?

01:06:07.980 | Well, they've been clear they don't want to.

01:06:09.920 | They have a proprietary investment, it's valuable,

01:06:11.880 | they have control of the ball.

01:06:13.440 | And in some ways you can say the same thing

01:06:15.840 | to any of the other big companies in the standards body,

01:06:18.120 | why doesn't Google standardize its stuff?

01:06:20.600 | - So you think it'll continue being

01:06:22.280 | like a kind of a dance partner to JavaScript,

01:06:24.640 | to the base JavaScript?

01:06:25.880 | - There's a hope that at some point,

01:06:27.640 | if they keep reconverging it and the standard

01:06:29.800 | doesn't break them and goes in a good direction,

01:06:32.040 | we will get at least the annotation syntax

01:06:34.800 | and some semantics around them.

01:06:37.040 | Because when you're talking about type annotations,

01:06:39.020 | they're generally on parameters and return values

01:06:42.360 | and variable declarations, they're cast operators.

01:06:45.560 | You want that syntax to be reserved

01:06:47.280 | and you want it to work the same in all engines.

01:06:50.000 | And this is where ideas like GILAD's

01:06:51.560 | pluggable type systems might be good,

01:06:54.240 | though then you could create the same problem

01:06:55.840 | you have with Lisp and Scheme,

01:06:56.960 | where there's a bunch of macro libraries

01:06:58.560 | and they don't agree and you have conflicts between them.

01:07:01.720 | But pluggable type systems could be one way to standardize.

01:07:04.960 | - What do you think about the giant ecosystem

01:07:07.040 | of frameworks in JavaScript?

01:07:09.840 | It feels like, because, I mean,

01:07:12.840 | this is a side effect of how many people use JavaScript,

01:07:16.040 | a lot of entrepreneurial spirit,

01:07:18.960 | like create their own JavaScript frameworks.

01:07:22.320 | And they're actually awesome in all different ways.

01:07:27.280 | And this is an interesting question

01:07:31.200 | about almost like philosophically

01:07:33.200 | about biological system and evolution,

01:07:35.280 | all that kind of stuff.

01:07:36.160 | Do you see that as good or should it,

01:07:38.600 | should some of them die out quicker?

01:07:40.400 | - I think that maybe they should.

01:07:41.880 | Now jQuery was a very clever thing.

01:07:44.840 | John Resig made this library that was sort of query and do

01:07:47.600 | and blended sort of CSS selector syntax

01:07:50.000 | with JavaScript sort of object graph or DOM querying

01:07:53.000 | and made it very easy for people to do things

01:07:54.920 | almost like they were learning jQuery as its own language,

01:07:57.680 | domain-specific language.

01:07:58.960 | And that I think reflected in part

01:08:02.080 | the difficulty of using the document object model,

01:08:04.960 | these APIs that were originally designed in the '90s

01:08:07.040 | for Java as well as JavaScript.

01:08:08.440 | They were very object-oriented or even procedural.

01:08:11.720 | They were very kind of verbose.

01:08:13.360 | And it took like a constructor call

01:08:15.440 | and three different hokey-pokey dances to do something,

01:08:18.400 | whereas in jQuery, it's just one line.

01:08:20.720 | So that fed back finally into the standards.

01:08:24.520 | It didn't mean we standardized jQuery.

01:08:25.960 | It wasn't quite that concise.

01:08:27.120 | But you find now with the modern standards

01:08:29.680 | that we were working on in the HTML5 sort of effort,

01:08:34.680 | that things became simpler,

01:08:35.840 | the fetch API and the query selector API,

01:08:38.840 | document.querySelector.

01:08:40.120 | A lot of things can be done now in raw JavaScript

01:08:42.560 | that you would make more concise and terse in jQuery,

01:08:46.640 | but it's not bad.

01:08:48.000 | It's pretty good.

01:08:48.840 | Whereas in the old DOM of 15 years ago,

01:08:51.120 | it was just too verbose.

01:08:51.960 | - So maybe the frameworks were born

01:08:53.800 | because JavaScript lacked some of the features of jQuery.

01:08:59.000 | And so now that JavaScript is swallowing what jQuery was,

01:09:03.520 | then the frameworks will,

01:09:05.240 | only the ones that truly add value will stick around

01:09:07.840 | and the other ones will die out.

01:09:09.000 | - And that highlights also this division

01:09:10.680 | between the core language JavaScript,

01:09:12.320 | which can show up in other places

01:09:13.600 | like Node.js on the server side

01:09:15.680 | and the browser specific APIs

01:09:17.680 | or the document object model APIs,

01:09:19.720 | which are even managed by the W3C,

01:09:21.560 | the standards body that was off in XML

01:09:23.800 | while I land on we were doing

01:09:25.320 | real JavaScript standards in ECMA.

01:09:26.960 | And you have this division of labor,

01:09:30.280 | division of responsibility and division of style

01:09:33.240 | and sort of aesthetics and also speed.

01:09:37.240 | So the document object model really stagnated

01:09:39.600 | after Microsoft kind of de-invested in the web.

01:09:43.560 | And Microsoft did something in their haste

01:09:45.880 | in the spirit of Netscape,

01:09:46.880 | doing things quickly and getting on first called DHTML.

01:09:49.760 | And some of their innovations

01:09:51.320 | that were like an alternative document object model

01:09:53.680 | didn't really get standardized until HTML5

01:09:56.840 | when we pragmatists set up for at the time,

01:10:00.040 | Ian Hickson, who went to Google, Apple and Mozilla said,

01:10:03.640 | "XML is not gonna replace HTML.

01:10:07.160 | HTML4 is too old.

01:10:08.720 | Let's standardize HTML5 based on all this good stuff,

01:10:12.160 | including that DHTML variant, dynamic HTML.

01:10:15.000 | - HTML5, it feels like to me,

01:10:17.240 | maybe you can correct me,

01:10:18.120 | like a beautiful piece of design work.

01:10:21.120 | It's not often with web stuff,

01:10:24.880 | you have this breath of just like,

01:10:27.320 | oh, whoever did this, this just feels good.

01:10:31.320 | What are your thoughts about HTML?

01:10:34.280 | Am I being too romantic?

01:10:35.840 | - A little bit, a little bit.

01:10:36.800 | - Are there flaws, fundamental flaws to it

01:10:38.680 | that I'm just not aware of?

01:10:40.080 | - My old friend Hicksey did a great job.

01:10:42.560 | He was another renegade physics student.

01:10:45.760 | And he was basically a QA guy at opera,

01:10:48.000 | but he obviously trained physics student

01:10:52.520 | and someone who could write, a Britisher.

01:10:54.720 | He developed test suites

01:10:57.840 | and he started thinking about them more axiomatically.

01:11:00.920 | Now this can be good because you can sort of systematize

01:11:03.960 | in a way that makes a better HTML,

01:11:06.000 | or you can get caught in the pragmatism of saying,

01:11:08.080 | well, we have to handle all of these edge cases,

01:11:09.840 | so we're just gonna have sort of a test matrix.

01:11:12.560 | And if the matrix is large,

01:11:14.000 | it will not be beautiful by many people's lights.

01:11:15.880 | Everyone likes to minimize along their preferred dimensions,

01:11:19.320 | the seven special forms and scheme or whatever.

01:11:21.680 | But reality is HTML needs to be big.

01:11:26.680 | It's kind of shambolic, it's a creative multi-paradigm.

01:11:30.400 | And Hicksey did a good job, I would say, with a bunch of it.

01:11:34.380 | Other people came in in the spirit of Ian Hickson

01:11:36.640 | to do HTML5 work and they've carried on that effort.

01:11:40.240 | And so it's a mix of pragmatism,

01:11:42.640 | de facto standards from the past being sort of combined

01:11:46.280 | or written down for the first time,

01:11:47.580 | and then rethought in a way that has a simpler syntax,

01:11:50.300 | like the fetch API instead of XMLHTTP request.

01:11:54.400 | - This video too as well,

01:11:55.440 | it ultimately, it feels like, maybe you can correct me,

01:11:58.080 | it feels like it was the nail in the coffin of Flash.

01:12:01.200 | - Steve Jobs saying no Flash on the iPhone,

01:12:03.040 | in my opinion, was the actual stake to the heart.

01:12:05.460 | But, well, I'm not sure what trope you wanna use.

01:12:09.160 | Flash was a zombie until just this year, right,

01:12:12.080 | or last year, I think last year was the end of Flash

01:12:14.160 | in main browsers.

01:12:15.240 | But Jobs really did the death blow.

01:12:18.280 | And yet, you're right, we had to make HTML5 competitive.

01:12:23.040 | I still don't think we got that beautiful

01:12:24.960 | timeline animation.

01:12:26.760 | - The timeline thing, so you like the time,

01:12:28.600 | I mean, me from, I used to animate all kinds of stuff

01:12:31.880 | inside Flash, plus there's a programming element.

01:12:34.920 | It was a little bit, I don't know if you can comment

01:12:38.220 | on that, but to me, it was a little bit like

01:12:39.860 | go-to statement, like in a sense that

01:12:42.640 | it was a little bit too chaotic.

01:12:44.900 | Like it didn't, that OCD part of me as a programmer

01:12:48.360 | wasn't satisfied by Flash.

01:12:50.180 | It feels like there was bugs that were introduced

01:12:52.760 | through the animation process that I couldn't debug easily.

01:12:55.340 | - Yes, I heard that too.

01:12:56.620 | I didn't use it, so I'm doing the grass is greener thing

01:12:59.260 | here.

01:13:00.100 | The thing I liked about the animation model

01:13:02.100 | was that it was this immutable function of time,

01:13:03.900 | so you could time warp and you could,

01:13:06.220 | if you dodged these bugs or worked carefully,

01:13:08.320 | you could really make it sing in ways that I think

01:13:10.660 | still a little challenging with web animation standards,

01:13:14.620 | but, or just using raw canvas and WebGL.

01:13:18.140 | But there's so many tools now that maybe it doesn't matter.

01:13:20.740 | And yet we had to do video, we had to do WebGL

01:13:25.740 | and then evolve it.

01:13:27.820 | We had to do web audio.

01:13:30.060 | But once we did all these things that helped Flash die,

01:13:34.140 | thanks to Steve Goves, we had something that

01:13:37.380 | people didn't realize.

01:13:39.580 | We had that vision that Mark and Jason had,

01:13:41.180 | this graphics capable to the metal, portable runtime.

01:13:46.180 | And we at Mozilla realized this and we saw JavaScript

01:13:51.600 | was something that you could compile to.

01:13:53.340 | Adobe had somebody in the Adobe labs doing this too.

01:13:55.620 | He had a project called Alchemy.

01:13:57.220 | We had somebody who's now at Google,

01:13:59.020 | Alon Zakai, who did his own LLVM based compiler

01:14:03.220 | that would take C or C++ and it would emit JavaScript.

01:14:06.140 | And you would think this is crazy.

01:14:07.620 | You're going from this sort of machine types,

01:14:09.820 | low level, controlled memory allocation language

01:14:13.080 | to this garbage collected dynamically typed,

01:14:16.340 | high level, higher level language.

01:14:18.020 | But Alon sort of just phenomenologically carved nature

01:14:22.420 | of the joint and found the forms

01:14:23.700 | that were fast in JavaScript.

01:14:25.580 | And then with Dave Herman, who I'd recruited

01:14:28.220 | from Northeastern University, who was a type theorist,

01:14:31.220 | and Luke Wagner, who's still at Mozilla,

01:14:33.800 | who was the compiler guy and the JIT guy,

01:14:36.820 | they figured out how to codify what Alon had done

01:14:40.180 | into a typed subset of JavaScript called Asm.js.

01:14:44.260 | And this is a strange thing to think about

01:14:45.660 | because it doesn't have new syntax.

01:14:47.260 | The types are casts that occur in dominator positions

01:14:52.260 | in the control flow graph.

01:14:54.140 | So it's like a hack on JavaScript and it's a subset.

01:14:56.740 | And it uses those bitwise operators

01:14:58.380 | that I talked about copying from Java

01:15:00.420 | to basically cast numeric types,

01:15:04.480 | which are double precision point into integers.

01:15:06.980 | And so inside JavaScript,

01:15:08.940 | in the kernel semantics are integers.

01:15:11.700 | And if you use these operators,

01:15:13.040 | if a compiler emits them in the right places,

01:15:15.580 | you can then treat them as typed values,

01:15:18.500 | typed memory locations, and you can type check your program.

01:15:22.380 | You can not only type check it, you can compile it.

01:15:24.940 | This is all in sort of linear time.

01:15:26.420 | Oh, and you can compile it

01:15:29.020 | to have deterministic performance.

01:15:30.420 | It doesn't touch the garbage collector.

01:15:32.580 | It calls a bunch of functions that come from the C functions

01:15:35.460 | or C++ code that you're compiling.

01:15:37.460 | And you can make the epic Unreal Engine

01:15:40.940 | go in 30 frames a second.

01:15:43.580 | And when we did this in 2013 in the fall,

01:15:46.860 | Tim Sweeney, I bet, didn't think it could be done quickly.

01:15:50.140 | He thought it would take years.

01:15:51.540 | And the team went to Raleigh to Epic,

01:15:54.060 | and in four days they had Unreal Engine ported

01:15:57.420 | by pressing a compile button.

01:15:59.940 | But they had to have WebGL, which came from OpenGL,

01:16:03.180 | ES came to OpenGL, which came from Silicon Graphics GL.

01:16:05.980 | They had to have Web Audio, so they could map OpenAL,

01:16:09.420 | which was another audio library standard,

01:16:11.000 | to Web Audio, which was kind of a Chrome idiosyncratic thing.

01:16:14.000 | But they could make it work.

01:16:16.260 | And they had to have Asm.js for fast C++ to JavaScript.

01:16:21.060 | And if you didn't have that fast compiler step,

01:16:25.220 | the JavaScript you'd write by hand

01:16:27.340 | trying to do an Unreal game would be too big and too slow.

01:16:29.780 | It would touch the garbage collector.

01:16:31.380 | It would not keep up with 30 frames a second

01:16:33.220 | on the 2013 hardware.

01:16:35.600 | So we demoed that at,

01:16:37.000 | this must have been fall 2012 now that I think about it,

01:16:40.420 | 'cause we demoed it at GDC, Game Developer Conference 2013,

01:16:44.420 | and people were stunned.

01:16:45.420 | That's like Unreal Engine, Unreal Tournament,

01:16:47.380 | running in my browser window.

01:16:48.860 | No plugin, no Flash, no Java, no--

01:16:52.660 | - So were those the early days of,

01:16:54.460 | because JavaScript now is able to run basically on par

01:16:58.380 | with a lot of the C++.

01:17:01.420 | - Yeah, and even before then,

01:17:02.820 | you had the fast JavaScript VMs in 2008 when Chrome came out.

01:17:06.560 | Just before it came out, Mozilla,

01:17:08.740 | my friend Andreas Gal and I and others

01:17:11.700 | hacked out TraceMonkey, our trace-based JIT.

01:17:14.460 | The SquirrelFishExtreme team at Apple did their JIT.

01:17:18.740 | And we were all competing

01:17:19.700 | on these crazy performance benchmarks.

01:17:21.900 | It was a little bit too much tuning of the benchmark.

01:17:23.540 | But JavaScript started getting fast,

01:17:25.220 | and developers started noticing it.

01:17:26.700 | But it was still kind of its own high-level language

01:17:29.780 | with garbage collection.

01:17:30.780 | The Asm.js step helped us go further,

01:17:33.260 | because until we really proved the concept,

01:17:37.100 | people were still saying, "Well, JavaScript's okay.

01:17:39.320 | "It's getting faster thanks to V8."

01:17:41.020 | Everybody gave Google credit, especially Google.

01:17:43.300 | But we need something to kill Flash.

01:17:44.940 | Let's use a portable native client code

01:17:47.300 | that Google had acquired, Native Client,

01:17:49.300 | which is a separate lineage for taking basically C code,

01:17:52.940 | compiling it into a software fault isolated container

01:17:56.260 | of some sort, using some kind of virtualization technique.

01:17:59.740 | And maybe it can even be in process

01:18:01.620 | and still be memory-safe.

01:18:02.540 | That would be awesome.

01:18:03.380 | But they ended up using process isolation, too,

01:18:04.780 | and that kind of weakened it.

01:18:06.300 | And in the end, it was like portable native client,

01:18:08.540 | okay, meet the new boss, same as the old boss.

01:18:11.060 | This is the Google Flash, right?

01:18:13.260 | But when we did Asm.js

01:18:15.060 | and we showed Unreal Engine working,

01:18:17.180 | I think it was only a matter of time

01:18:18.660 | before Google threw in the towel.

01:18:19.820 | And in fact, everybody agreed in spring of 2015,

01:18:23.140 | we're gonna take what was proven by Asm.js

01:18:25.500 | and make a new syntax, a binary syntax that's efficient,

01:18:29.460 | that loads into the same JavaScript VM

01:18:31.580 | that JavaScript loads into.

01:18:32.620 | So there'll be two source languages,

01:18:34.100 | one VM, very important, one garbage collector,

01:18:36.220 | one memory manager, one set of compiler stages.

01:18:40.800 | And that's called WebAssembly.

01:18:42.300 | And that's the successor to Asm.js.

01:18:44.260 | And it's important that it have binary syntax

01:18:46.340 | because at the end of the day, especially on mobile,

01:18:48.780 | if you're downloading JavaScript,

01:18:50.100 | even if you're using LZ compression on the wire,

01:18:52.040 | that's cool, but you've got to blow it out into memory

01:18:53.900 | and then parse the silly eight character function keyword

01:18:56.860 | that I picked when I should have used something shorter.

01:19:00.780 | I picked it because of awk, the Unix tool.

01:19:03.300 | So anyways.

01:19:05.380 | - I'm not following, I want to,

01:19:06.220 | but I'm not following the awk thread.

01:19:07.940 | - Yeah, don't worry about it.

01:19:09.780 | - Is it surprising to you that

01:19:12.260 | how damn fast JavaScript is these days?

01:19:14.660 | I mean, 'cause you've been through the whole journey.

01:19:16.820 | I know every step of the way, but is it like,

01:19:19.660 | I mean, it feels incredible.

01:19:22.140 | - It does, but I knew, so the funny thing is,

01:19:24.220 | computer science is this big karmic wheel, right?

01:19:26.820 | Wheel of Fortuna.

01:19:27.900 | And in the, it was about the '97,

01:19:32.420 | I was loaned by Netscape to do due diligence for Sun

01:19:37.060 | in their acquisition of Animorphic,

01:19:39.260 | which was David Unger and friends,

01:19:43.860 | people, Craig, I'm forgetting his name,

01:19:46.220 | he went to Microsoft.

01:19:47.660 | These Stanford language buffs who had taken Smalltalk

01:19:51.140 | and then David created Self

01:19:53.460 | as a simpler sort of Smalltalk language

01:19:56.220 | and made really fast, just-in-time compiling VMs for them.

01:20:00.580 | And they, well ahead of Java Hotspot or JavaScript V8

01:20:05.580 | or any of these modern VMs,

01:20:07.620 | figured out how to make dynamic code fast

01:20:10.740 | because Smalltalk is dynamic language, right?

01:20:13.060 | It has classes, it has, I think,

01:20:15.060 | more lockdown declarative syntax than JavaScript,

01:20:17.380 | but it's fundamentally dynamic, you don't declare the types.

01:20:20.980 | But you could infer the types as the program runs

01:20:24.140 | and you start to form these ideas

01:20:26.080 | about what types are actually flowing

01:20:27.600 | through key operations and you form little

01:20:30.580 | so-called polymorphic inline caches

01:20:32.720 | that are optimized machine code.

01:20:35.660 | The cache is the machine code that assumes,

01:20:38.300 | does a quick check to make sure the type is right,

01:20:40.140 | and if it's not right, it bails to the interpreter.

01:20:42.380 | If it is right, you go pretty fast.

01:20:44.500 | And that short test is a predicted branch,

01:20:47.100 | so things are pretty quick.

01:20:48.620 | All that amazing stuff I knew about in the '90s

01:20:51.340 | and I didn't have time to do it

01:20:53.780 | and Anamorphic got bought by Sun and they did Hotspot.

01:20:57.220 | And you needed that even in Java because at scale,

01:21:00.060 | Java has some dynamic aspects due to invoke interface.

01:21:04.340 | You can have basically collections of Java code

01:21:06.720 | where you don't know at the time each module

01:21:10.900 | or package is compiled exactly what's being called,

01:21:13.900 | what subclass or what implementation

01:21:16.300 | of an interface is being called.

01:21:17.820 | And so you want to optimize using this sort

01:21:19.580 | of dynamic polymorphic caching there too.

01:21:21.740 | And they did that and Hotspot is an amazing beast.

01:21:24.620 | I've met like 13 people who all claim they created it.

01:21:27.740 | I think one of them may deserve credit more than others.

01:21:30.540 | But I didn't get to do that in JavaScript

01:21:34.660 | and when we knew that Google was going

01:21:38.060 | to do their own browser, which we knew at Mozilla around 2006

01:21:41.060 | I also met the team that did V8.

01:21:45.180 | And it turns out it was Lars Bach

01:21:46.620 | who was one of the young engineers from Anamorphic

01:21:49.300 | who got acquired by Sun.

01:21:50.660 | And so Lars is like one of the world's expert

01:21:52.940 | on these kinds of virtual machines

01:21:54.940 | and he picked my brains about JavaScript.

01:21:56.780 | I could tell he didn't like it at the time

01:21:58.100 | but he had to do it.

01:22:00.460 | - Oh, really, interesting.

01:22:01.380 | - Yeah, in 2006 launch at Google's campus.

01:22:03.620 | And then I had another friend who was DevRel at Chrome

01:22:07.020 | and he said, yeah, we don't know what they're doing.

01:22:08.300 | This is getting 2007 to fall, getting toward 2008.

01:22:11.980 | We're trying to get Chrome out

01:22:12.980 | and we don't know what's going on with the V8 team.

01:22:14.460 | They're off in Aarhus, Denmark,

01:22:16.860 | rewriting their engine four times, which is good.

01:22:19.260 | That's the right way to do this kind of development.

01:22:22.260 | They were learning JavaScript, including all its quirks

01:22:24.260 | which they came to hate, the fire of a thousand suns

01:22:26.980 | which is one of the reasons that Lars and company

01:22:28.860 | did Dart, their own language.

01:22:30.900 | But they also made the language fast.

01:22:32.860 | And meanwhile, we knew this was happening.

01:22:35.220 | So we got our act together with TraceMonkey,

01:22:38.460 | our tracing JIT at Mozilla.

01:22:40.060 | And Apple, I think, was also aware.

01:22:41.820 | And so they were doing their own JIT.

01:22:42.980 | So the era of JIT-ed fast JavaScript in 2008

01:22:46.420 | had this prehistory going back to Smalltalk itself

01:22:49.540 | and anamorphic.

01:22:50.940 | And again, the lineage is interesting

01:22:53.060 | 'cause you had Lars at anamorphic

01:22:54.460 | and then he ends up at Google.

01:22:55.940 | - Yeah, and today we have an incredibly fast language

01:23:00.140 | that like you said, still, without hate, you can't have love.

01:23:04.740 | So I think there's both love and hate for this dance,

01:23:09.740 | this rich, complex dance of JavaScript

01:23:12.060 | throughout its history.

01:23:12.900 | - There's a dialectic for sure.

01:23:15.300 | - Today, JavaScript is the most popular language

01:23:18.940 | in the world.

01:23:20.120 | Why, by many measures, why do you think that is?

01:23:24.220 | Is there some fundamental ideas

01:23:28.420 | that you've already spoke to a little bit

01:23:29.700 | but sort of broader that you think

01:23:31.380 | is the most popular language in the world?

01:23:32.820 | - So I think I did, by doing first class functions

01:23:36.220 | and taking the good parts of the C operator hierarchy

01:23:40.780 | and just keeping things simple enough,

01:23:44.060 | maybe it could have been simpler,

01:23:45.060 | but I had to make it look like Java

01:23:47.340 | and interoperate with Java,

01:23:48.740 | that there was inherent goodness,

01:23:52.820 | Aristotelian quality there.

01:23:54.380 | And people perceive that

01:23:56.060 | even through all the quirks and warts.

01:23:57.700 | And then over time, working on it with the standards body,

01:23:59.900 | working on it not only as a core language,

01:24:02.760 | but in the context of HTML5 and making the browser better,

01:24:06.620 | listening to developers, thinking about,

01:24:08.940 | this is something that Nick Thompson wrote nicely about

01:24:11.460 | on the Hacker News, I was very flattered.

01:24:12.580 | He said, "Java was this thing

01:24:14.220 | "where the experts were writing the code

01:24:15.900 | "and it was compiled and you had to declare all your types.

01:24:17.920 | "And Sun didn't really give a damn about

01:24:20.860 | "the average programmer who wanted to build real web apps,

01:24:23.720 | "dynamic things."

01:24:25.020 | And I was in there meanwhile doing a bunch of people's jobs,

01:24:28.800 | making JavaScript survive those early years

01:24:31.540 | when it was kind of touch and go, right?

01:24:33.580 | JavaScript was considered a Mickey Mouse language.

01:24:36.100 | It was for annoyances like the scrolling text

01:24:38.620 | at the bottom of the browser in the status bar.

01:24:40.780 | But I kept listening to developers,

01:24:42.440 | working with them and trying to make it run

01:24:45.420 | in that single threaded event loop in a useful way.

01:24:47.680 | And I think that forged something

01:24:49.840 | that people have come to love.

01:24:51.160 | Now you don't always love the best thing, right?

01:24:53.660 | I talked about, Shakespeare,

01:24:56.060 | sunnet about, "I'm Mr. Sizer, nothing like the sun."

01:25:00.140 | Or the scene from Josh Whedon's film,

01:25:03.780 | Serenity at the end where the actual piece in the score

01:25:07.700 | by David Newman is called Love,

01:25:09.020 | where Captain Mal is teaching River Tam

01:25:12.820 | about how to pilot a ship.

01:25:13.840 | And she's a super genius, super soldier.

01:25:15.340 | She knows how to do it already.

01:25:17.100 | And he's basically talking about how you have to love

01:25:19.460 | the ship because if you don't, it's gonna kill you.

01:25:22.020 | And then the piece falls off the ship.

01:25:23.820 | It's kind of like JavaScript.

01:25:25.220 | You have to love it.

01:25:26.900 | You have to love it because now people say,

01:25:28.920 | "Ah, we're stuck with it

01:25:30.260 | 'cause it got this priority of place."

01:25:31.620 | - But there's love underpinning that.

01:25:33.320 | And actually, listening to developers,

01:25:35.860 | that's kind of beautiful.

01:25:36.700 | There's most successful products in this world

01:25:41.100 | with all the messes, with all the flaws.

01:25:43.060 | Perhaps the flaws themselves are actual features,

01:25:45.320 | but that's a whole nother, that's a discussion about love.

01:25:48.140 | But underneath it, there's something

01:25:50.180 | that just connects with people.

01:25:52.380 | - And it has to keep connecting.

01:25:53.440 | If JavaScript kind of went off in this,

01:25:55.620 | people sometimes complain about ES6.

01:25:57.300 | Oh, you put classes in JavaScript.

01:25:58.660 | I hate classes.

01:25:59.500 | You've ruined it.

01:26:00.400 | But it's not true.

01:26:01.380 | It's a dynamic language.

01:26:02.620 | Smalltalk had classes.

01:26:04.780 | Python has classes.

01:26:06.040 | There are lots of Lisp variants that had classy systems.

01:26:10.080 | Common Lisp.

01:26:12.700 | So people who don't reject it based on some sort

01:26:16.740 | of fashion judgment do use it

01:26:21.020 | and do interact with the standards body.

01:26:23.340 | The standards body is competing browser vendors mainly,

01:26:25.800 | but also now big companies that use JavaScript heavily

01:26:28.440 | the PayPal's and other such companies, Salesforce.

01:26:32.940 | And they have to cater to web developers.

01:26:36.260 | They have to hire developers who know JavaScript.

01:26:38.980 | They have to keep their engines up to the latest standard.

01:26:43.060 | And this creates all this sort of social structure

01:26:45.860 | around JavaScript that is unusual.

01:26:47.780 | I mean, you get C++ buffs that follow the inner workings

01:26:51.380 | of C++, what is it now?

01:26:54.420 | 21 something.

01:26:55.900 | I don't know.

01:26:56.740 | I've lost track.

01:26:57.560 | But it's a more rarified group.

01:26:58.980 | It's more like the old language, gray hairs.

01:27:02.820 | Whereas JavaScript is a younger

01:27:04.500 | and more vibrant and large crowd.

01:27:07.380 | - There's a community feel to it.

01:27:09.260 | There's a echoes, perhaps I don't wanna draw

01:27:12.180 | too many similarities.

01:27:13.940 | Maybe you can comment on it.

01:27:15.420 | There's a C++ is like Wall Street

01:27:18.980 | and the JavaScript is like Wall Street bets

01:27:21.220 | from the recent events.

01:27:22.620 | It's like, there's a chaotic community of all.

01:27:24.940 | And there's some power from that distributed crowd

01:27:28.800 | of people that ultimately--

01:27:30.240 | - It's more demotic.

01:27:31.120 | It's more of the people.

01:27:32.640 | It lets people in without requiring these credentials.

01:27:35.840 | I remember in the late 90s into the noughties,

01:27:38.400 | people were all getting Java credentials.

01:27:39.880 | And I knew people and friends knew people

01:27:42.640 | who became Java programmers.

01:27:44.120 | And you knew they really should have been

01:27:45.320 | like nature guides or pilots.

01:27:47.200 | They hated programming, but they thought,

01:27:49.000 | I gotta make money.

01:27:49.840 | I'm gonna become a Java programmer.

01:27:51.740 | - Do you have some, because it's such a monumental moment

01:27:54.920 | in our current history as a quick aside,

01:27:58.080 | do you have thoughts about this huge distributed,

01:28:02.200 | crowdsourced financial happenings with Wall Street bets?

01:28:06.600 | - Yeah.

01:28:07.440 | - That's like nobody could have,

01:28:08.500 | well, you could have predicted,

01:28:09.700 | but the scale and the impact of this kind

01:28:12.080 | of emergent behavior from independent parties

01:28:15.000 | that could happen.

01:28:15.840 | - Like I said, my own experience with the dismal science

01:28:19.900 | as with physics led me to reject a lot of bad models.

01:28:22.520 | And economics was always compromised by politics,

01:28:26.280 | political economy.

01:28:27.400 | You could also argue that it was,

01:28:29.480 | it used to be a branch of moral philosophy.

01:28:31.020 | So it was concerned with the good and it became divorced

01:28:33.960 | and became sort of in this quasi Newtonian way,

01:28:36.600 | just about everything's just running by itself.

01:28:39.160 | Don't worry about it.

01:28:40.280 | This monopoly is crushing your Netscape company,

01:28:42.080 | but that's just nature.

01:28:43.760 | - And economics couldn't,

01:28:45.200 | or doesn't really have good models

01:28:46.620 | for the Wall Street bets subreddit.

01:28:49.000 | - They know how to squeeze a short, right?

01:28:52.320 | So the amazing thing is you have Robinhood app,

01:28:56.480 | which was again, supposedly for the demos, for the people,

01:29:00.760 | and eliminated the fee through various kinds of straddles

01:29:05.160 | or some kind of spread operation

01:29:09.240 | that helped them eliminate the fee or eat the fee.

01:29:11.880 | And in fact, as a broker in these days,

01:29:14.560 | because it takes two days to settle,

01:29:15.840 | there's counterparty risk as they found out.

01:29:18.640 | And so the Wall Street bets people,

01:29:23.400 | the memes are like the Terminator robot

01:29:25.700 | with $600 STEMI check and the hedge funds

01:29:30.160 | the little girl hiding under the desk.

01:29:32.400 | There is a problem, which I talked about

01:29:36.040 | in the recent podcast, which I'm conscious of

01:29:38.280 | from the history of the web.

01:29:39.320 | And that is, you could say it's monopoly,

01:29:41.240 | which antitrust wasn't enforced

01:29:42.960 | after US Microsoft for a long time.

01:29:45.120 | And a lot of this was due to the money interests

01:29:48.000 | buying control of politicians.

01:29:50.880 | And in Plato's five regimes, that's oligarchy.

01:29:54.400 | That's where we are.

01:29:56.160 | And now we're seeing a fight against the oligarchs.

01:29:58.720 | I don't know if it'll work,

01:29:59.560 | but you're definitely seeing it.

01:30:00.760 | And it's also kind of hackerish, right?

01:30:02.360 | It's got a hacker ethos.

01:30:04.320 | Hey, Robinhood, no fees.

01:30:05.640 | Oh, interesting.

01:30:06.460 | Hey, I can buy a fraction of a share in this thing,

01:30:09.560 | or I can keep buying with my stimulus check.

01:30:12.240 | So I mentioned Hegel seeing Napoleon on the horse.

01:30:16.920 | Hegel also talked about the cunning of reason,

01:30:19.160 | that you have this sort of God sees history in full,

01:30:23.040 | and if you believe in God, or we don't know the future,

01:30:25.880 | but there's always this sort of fly in the ointment,

01:30:28.320 | this unintended consequence that confounds

01:30:32.080 | the best plans of the powers that be.

01:30:35.600 | And we're living through it.

01:30:37.000 | I'm glad it's not street warfare or mechanized warfare,

01:30:41.480 | 'cause it has been in the past.

01:30:44.760 | It's more like soft power.

01:30:47.120 | - Yes.

01:30:47.960 | - And people are fighting back.

01:30:49.360 | - Do you think it's possible,

01:30:51.760 | so JavaScript used to be for the front end of the web.

01:30:56.760 | It's now increasingly so being used for backend,

01:31:01.600 | like running stuff that's like behind the scenes.

01:31:06.080 | And it's also starting to be used quite a bit

01:31:10.260 | for things like TensorFlow.js.

01:31:12.240 | So starting to actually use these heavy duty applications

01:31:16.040 | that are using neural networks, machine learning,

01:31:17.680 | and so on in the browser.

01:31:19.380 | Is it possible in 10, 20, 30 years

01:31:24.120 | that basically most of the world runs on JavaScript?

01:31:28.640 | - This is a dystopia and a nightmare to some people.

01:31:31.240 | (Dave laughs)

01:31:33.160 | When we did Asm.js and WebAssembly,

01:31:34.920 | I would joke and meme people with scenes like

01:31:37.360 | Neo waking up in his pod in the Matrix,

01:31:39.520 | and he's all skinny and weak and hairless.

01:31:41.920 | And you realize in the future that you're living

01:31:46.720 | in some simulation that it's all running on JavaScript,

01:31:49.160 | and you just scream forever.

01:31:50.600 | It's possible.

01:31:52.840 | Gary Bernhardt does these funny talks.

01:31:54.840 | He did Wat.js, and then he did this

01:31:56.880 | Life and Death of JavaScript, I think it's called,

01:31:59.560 | where he took some clever ideas

01:32:02.000 | that actually have a thread of credibility to them.

01:32:05.320 | But I mentioned software fault isolation.

01:32:09.200 | In the old days when we were using computers,

01:32:10.600 | we said we're gonna use the Unix monolithic monitor,

01:32:14.120 | and it's the privileged program.

01:32:15.980 | This is before you even had hardware rings of protection.

01:32:18.400 | Those, some of the early '60s operating systems

01:32:20.440 | used hardware protection zones.

01:32:22.680 | But Unix is privileged, and the program

01:32:25.920 | that runs user code in a process is hosted.

01:32:30.440 | It's the guest in the host, and you get to suspend it.

01:32:33.640 | You get to kill it.

01:32:35.240 | If it crashes, it doesn't take down the whole OS.

01:32:37.280 | It's a wonderful idea.

01:32:39.360 | But the call into the kernel is expensive,

01:32:42.640 | the system call, so-called.

01:32:43.800 | And this has even been optimized now

01:32:45.160 | for things like getting the time of day

01:32:46.300 | so it doesn't actually enter the kernel.

01:32:48.300 | And meanwhile, hardware architectures

01:32:52.760 | and virtualization techniques have gone

01:32:54.420 | in a different direction, even to the point

01:32:56.360 | where you can do software fault isolation very cheaply

01:32:58.800 | without entering the operating system kernel.

01:33:00.720 | And so you get unikernels and exokernels

01:33:02.880 | and very lightweight VMs.

01:33:04.320 | And so Gary took this idea and said,

01:33:07.520 | JavaScript will take over computing

01:33:09.160 | because the system call boundary's too expensive,

01:33:11.640 | so everything ends up in JavaScript

01:33:12.920 | with these lighter-weight isolation enforcement mechanisms.

01:33:16.880 | It's not totally beyond belief.

01:33:18.560 | It'd be WebAssembly, too.

01:33:20.880 | - It's nice to ask you sort of for advice.

01:33:25.280 | There's so many people that are interested

01:33:26.920 | in starting to learning about programming,

01:33:29.120 | getting into this world.

01:33:30.320 | Is there some number of languages,

01:33:34.960 | three to five programming languages

01:33:36.440 | that you would recommend people learn,

01:33:38.200 | or maybe a broader advice

01:33:41.680 | on how to get started in programming?

01:33:43.920 | - Well, so you asked about machine learning,

01:33:45.360 | and JavaScript is a general-purpose language,

01:33:47.800 | and it's a language that's not that great

01:33:52.800 | for doing matrix operations

01:33:56.260 | or doing parallel programming, I would say,

01:33:59.600 | without using some extensions or some libraries

01:34:02.080 | that have some magic in them.

01:34:03.560 | So if someone wanted to learn,

01:34:06.720 | there are amazing languages in sort of the APL family

01:34:11.040 | that are very useful for, I would say, linear algebra,

01:34:14.000 | which gets to a lot of the kernels in machine learning.

01:34:16.760 | And so APL had like J and then K,

01:34:19.940 | and they're K variants

01:34:22.160 | because the guy that did K is still going,

01:34:24.600 | and they're proprietary, but he's still innovating there.

01:34:27.480 | Python is used.

01:34:30.360 | So people talk about TensorFlow.js.

01:34:31.680 | Well, it's not that surprising

01:34:33.400 | 'cause Python was heavily used for machine learning,

01:34:36.320 | and Python was always, you know,

01:34:38.680 | they didn't have this fast just-in-time compiler tradition.

01:34:41.480 | There were some projects that tried this,

01:34:43.120 | and some of them were interesting.

01:34:45.280 | PyPy was interesting.

01:34:46.240 | But the philosophy with Python was,

01:34:49.960 | oh, you need to go fast, write a C plugin,

01:34:51.720 | and drop into C code.

01:34:53.040 | So I think people should look at multiple languages

01:34:57.480 | because there are different tools in the belt.

01:34:59.480 | If you're trying to do supervision or rapid prototyping,

01:35:02.800 | you want a dynamic language.

01:35:03.920 | You wanna throw things together and see what works.

01:35:06.520 | If you are trying to go down to the metal very fast,

01:35:09.280 | well, I'm an old C hacker,

01:35:10.720 | but I was also the executive sponsor of Rust at Mozilla,

01:35:14.280 | and Rust has now escaped from that sort of nest

01:35:17.880 | where it was born to be adopted by a bunch of companies

01:35:21.440 | to have a foundation in the works.

01:35:23.200 | Some of the key core team members

01:35:24.760 | are working now at Amazon and other places.

01:35:27.660 | So it looks like Rust has reached escape velocity.

01:35:30.080 | And Rust is an interesting language

01:35:31.360 | because one of our goals there,

01:35:33.040 | one of the reasons I sponsored it was,

01:35:34.480 | we were all tired of seeing

01:35:35.640 | those remote code execution vulnerabilities

01:35:38.080 | due to C and C++.

01:35:40.160 | And we thought, can we have a sort of safety property

01:35:43.800 | through a type and effect system or an ownership system?

01:35:46.880 | And Rust has that.

01:35:48.280 | And that ownership system is interesting

01:35:49.960 | because it doesn't just give you memory safety.

01:35:52.120 | There's a sort of theorem for free,

01:35:53.460 | a dual that falls out for protection against data races.

01:35:57.160 | So Rust is better for low-level programming.

01:35:59.720 | You delimit your unsafe code where you do have to be unsafe,

01:36:02.880 | and you can prove certain facts about memory safety

01:36:05.280 | and race condition avoidance.

01:36:08.800 | And so I think people should learn these new languages.

01:36:11.320 | I think Go is a great language.

01:36:13.320 | I admire the Unix people who did that.

01:36:16.480 | Ken Stoll was involved, Rob Pike, of course,

01:36:19.040 | David, what's his name, and other people.

01:36:23.400 | Go is a huge success, really, on the server side,

01:36:27.080 | mainly where you have a lot of networking to do.

01:36:30.120 | And it's garbage collected, but it's also very pragmatic.

01:36:33.800 | It has that sort of C flavor.

01:36:35.400 | As an old C hacker, I can't get used to the fact

01:36:37.040 | that they swapped the type and declarator

01:36:39.600 | in the declaration order.

01:36:41.160 | - I haven't used Rust, but this is one of the most respected

01:36:44.200 | and loved languages currently, so it's interesting.

01:36:46.280 | - Yeah, and it's still young.

01:36:47.400 | You look at these things,

01:36:48.400 | JavaScript is now considered old.

01:36:50.140 | - It's gone through so many versions

01:36:52.520 | that you can fall in love with it all over again.

01:36:54.880 | - 25 plus years, it's an adult.

01:36:57.760 | (laughing)

01:36:58.760 | It should be out of the house.

01:36:59.800 | But it could be around another 25 years.

01:37:03.420 | Cannot rule it out.

01:37:04.260 | So Rust could be around for a long time.

01:37:06.280 | The longer you're around, the more likely you're Lindy

01:37:08.920 | and you're around your wife.

01:37:10.400 | - A lot of people ask me,

01:37:11.600 | I'm often torn between recommending either Python

01:37:15.800 | or JavaScript as the first language to play with,

01:37:18.080 | because it's difficult,

01:37:20.880 | because it's so easy to do JavaScript incorrectly.

01:37:24.840 | It's much easier to do it correctly these days,

01:37:29.440 | or like well, like learn about programming.

01:37:31.640 | But the cool thing about JavaScript

01:37:33.940 | is that you can create stuff

01:37:36.120 | that will put a smile on your face.

01:37:38.120 | Like as a developer, you can create stuff

01:37:41.440 | and it'll visually look like something and it'll do stuff.

01:37:44.200 | And it makes you feel good.

01:37:45.720 | It makes you fall in love with programming.

01:37:47.320 | With Python, you could do the same.

01:37:48.700 | It's a little slower.

01:37:49.540 | And with C++, it takes five to 10 years

01:37:51.560 | to write a program that actually does something.

01:37:54.100 | So like there's that tension between,

01:37:56.800 | is JavaScript the right first step or is it Python?

01:37:59.640 | And I've been going back and forth on those two.

01:38:01.880 | - I have my Python, right?

01:38:02.760 | It came from a lineage of ABC,

01:38:05.280 | which was a pedagogical language in the Netherlands.

01:38:08.600 | And it was a good teaching language too.

01:38:14.560 | I think it is a good teaching language.

01:38:17.980 | And it's a little more restrictive

01:38:19.120 | in that if you misspell something

01:38:21.960 | in a way that JavaScript might let run,

01:38:25.200 | let reach run time, it'll get stopped

01:38:26.880 | at syntax check in Python.

01:38:30.600 | That's good for beginners.

01:38:31.700 | I think the sloppiness that some people object to,

01:38:34.760 | like people were just tweeting at me,

01:38:37.160 | having just learned JavaScript.

01:38:38.640 | They said, "I can take a number and I can index into it

01:38:41.120 | "and get undefined out of it as a property."

01:38:42.880 | And why is that?

01:38:43.800 | A number is not an object.

01:38:44.840 | And I explained why it is.

01:38:46.200 | 'Cause like in Java, the primitive types,

01:38:48.340 | which unfortunately are not objects,

01:38:49.840 | can be automatically boxed or wrapped by an object.

01:38:53.260 | And I made that implicit.

01:38:55.440 | In Java, it's typed and you have to declare things

01:38:58.220 | and you'll get type errors.

01:39:00.480 | But there are cases in Java where you get auto-boxing

01:39:03.080 | or auto-wrapping 'cause you've declared that you want it.

01:39:06.020 | In JavaScript, it just happens.

01:39:07.540 | And so once I explained it, I'm like, "Oh, wow, I get it."

01:39:10.300 | But it also means that you can commit a blunder that just--

01:39:13.640 | - You don't get punished for it, you don't detect.

01:39:15.340 | - You get an undefined value

01:39:16.800 | and you don't know where it came from.

01:39:17.840 | - Right.

01:39:18.680 | I've been reading a lot about military history recently.

01:39:23.360 | And one way to paint the picture of browsers,

01:39:26.760 | internet browsers, is through the various wars

01:39:29.280 | throughout its history.

01:39:30.620 | I don't know if that's a useful way to look at it,

01:39:32.300 | but we've already talked a little bit about Netscape

01:39:34.660 | and Internet Explorer in the early days.

01:39:38.400 | Can you tell the story of the different wars,

01:39:41.260 | if that's at all an interesting way to look at it,

01:39:43.440 | of the '90s and to today?

01:39:46.460 | - Yeah, so I mentioned that Microsoft,

01:39:49.120 | which was convicted for it, did abuse its monopoly,

01:39:51.520 | but they had a pretty good team by the time they did IE4.

01:39:55.120 | And Netscape, unfortunately, I was like second floor

01:39:57.920 | and I was friends with all the first floor people,

01:39:59.520 | the front end guys who did the JavaScript event hookup

01:40:02.720 | and things like that, that team was fairly burnt out.

01:40:07.560 | And I think having gone public,

01:40:10.880 | the upper management wanted to buy a bunch of companies

01:40:13.320 | to try to go head to head with Microsoft.

01:40:15.360 | Didn't work, but buying a bunch of companies

01:40:17.760 | usually doesn't work.

01:40:19.360 | I think the modern sort of approach roughly

01:40:22.800 | is like Mark Zuckerberg took,

01:40:25.280 | which is to keep them at arm's length

01:40:26.880 | and let them do their thing.

01:40:28.320 | And now that he's pulling WhatsApp in

01:40:29.920 | and people are fleeing it

01:40:31.440 | because it's tied into the ad surveillance.

01:40:33.480 | But for a while they were keeping it separate.

01:40:36.920 | It really does work 'cause you bought it for its value,

01:40:39.000 | it's complimentary and you're not messing with it.

01:40:41.080 | With Netscape, when they bought a bunch of companies,

01:40:43.560 | they had some of the first floor people

01:40:45.440 | or the founders burned out.

01:40:46.920 | They had newcomers who wanted their turn to do the browser

01:40:51.400 | and they hadn't really done browsers or understood them.

01:40:54.400 | And so Netscape 4 was originally supposed to be 3

01:40:57.320 | and it was so late, they renumbered it.

01:40:58.800 | We did a 3 release, Jamie and a few others

01:41:01.400 | put some extra effort into secure MIME

01:41:03.640 | was supported in the mail, built-in mail program.

01:41:07.960 | And Netscape 4 was late and it was only on Windows at first

01:41:11.840 | and Microsoft had really started doing better.

01:41:14.880 | Like they do, they copy and the first version is trash

01:41:17.280 | and the second one, you're starting to feel threatened.

01:41:19.040 | The third one, you can tell what's gonna happen

01:41:21.400 | and the fourth one's good.

01:41:22.440 | - And plus there's the benefit, like you said,

01:41:24.200 | that it comes as a default browser.

01:41:26.080 | - Yes, and yet Netscape's screwing it up

01:41:28.760 | and Microsoft really putting some quality people on it.

01:41:32.520 | IE 4 was good, on Windows it was good.

01:41:34.200 | And they did the dynamic HTML innovations,

01:41:37.040 | they, Scott Isaacs, my old buddy,

01:41:39.680 | a former accountant who programmed in BASIC

01:41:42.360 | and became what Microsoft calls a program manager,

01:41:45.520 | which is kind of an elevated position.

01:41:47.000 | It's, you can be a programmer or an engineer track

01:41:50.520 | but you switch to it and you sort of lead a lot of design

01:41:53.440 | and standards efforts.

01:41:54.680 | And so Scott Isaac put in a lot of those funky DHTML APIs

01:41:58.920 | that didn't quite have the same flavor

01:42:01.000 | as the stuff that I did.

01:42:02.520 | And neither of them was like the later

01:42:04.200 | sort of verbose Java like DOM W3C standardized.

01:42:08.560 | But IE 4 was pretty darn good.

01:42:10.560 | I remember a friend, Scott Furman and I

01:42:12.480 | got invited by Scott Isaacs to Gordon Beers in San Jose.

01:42:16.680 | They were doing a preview of IE 4, this must have been 1997.

01:42:21.520 | And Scott said, yeah, we've got,

01:42:24.120 | here's the new graphics stuff we're doing.

01:42:25.560 | We've got something like your Netscape layers.

01:42:28.240 | We've got VML, a vector markup language.

01:42:31.800 | We can do like virtual reality

01:42:33.440 | and Scott and I looked at each other

01:42:34.920 | and said, we're doomed, right?

01:42:36.840 | Microsoft was starting to fire on all cylinders.

01:42:38.800 | So I have to give them credit for that

01:42:40.440 | even though they abused their market power.

01:42:42.960 | And maybe I shouldn't give them credit

01:42:44.320 | for having the resource to hire talented people

01:42:45.920 | but they did a credible job on IE 4.

01:42:48.880 | What really was bad was that phase of the browser wars

01:42:51.920 | ended with monopoly and perhaps due to the antitrust case,

01:42:56.080 | perhaps due to regulation in Europe,

01:42:58.140 | perhaps just due to Microsoft

01:43:00.560 | not liking dealing with standardization.

01:43:03.120 | They let it rot.

01:43:05.440 | They just abandoned it, IE 5, 5.5, IE 6 later,

01:43:10.440 | but these were not well-maintained.

01:43:12.320 | They had a lot of security bugs.

01:43:14.400 | - And it was really closed and outdated too,

01:43:16.680 | even though it was getting updated, it's just weird.

01:43:19.040 | - Browsers like Mozilla and then Firefox were adding tabs.

01:43:22.120 | Opera had a version of tabs and they didn't add tabs

01:43:25.960 | and they pop up blocking,

01:43:27.560 | something I should have done from the start.

01:43:28.840 | People realized that you can tell

01:43:30.160 | when the user clicks something

01:43:31.520 | and it goes in JavaScript to open a little window

01:43:34.280 | that you can sort of inspect the stack

01:43:36.200 | and see that the click originated that

01:43:38.120 | and it's probably okay.

01:43:39.400 | Whereas if you're just loading a script

01:43:40.800 | and it opens a new window, that's a spam technique

01:43:43.840 | and you should block it.

01:43:45.180 | - Tabs were a brilliant innovation.

01:43:47.360 | Like you said, Opera had it,

01:43:48.360 | but I remember I fully switched to Firefox the moment.

01:43:52.040 | I remember the moments of first using tabs in Firefox

01:43:57.040 | and not liking it for the first few minutes

01:44:00.160 | and then like, wait a minute.

01:44:02.400 | - You get the groove, yeah.

01:44:03.240 | - You get the groove and you understand.

01:44:05.520 | So that timing, what year was this?

01:44:08.520 | Because also as a aspiring web designer, I use table.

01:44:13.320 | So we didn't mention layout or CSS much.

01:44:16.240 | There's also a change in the way

01:44:18.520 | like the frames were going away.

01:44:21.040 | So there's a change in the way websites looked

01:44:23.600 | and behaved and all that kind of stuff.

01:44:25.240 | - CSS finally, which Microsoft embraced with IE4

01:44:28.400 | and Netscape never really did, right?

01:44:30.680 | CSS became a better standard over time

01:44:33.440 | for doing table layout that relieved you of the need

01:44:36.600 | to use what are called spacer GIFs, spacer GIFs, right?

01:44:40.280 | Images you would throw into space out tables.

01:44:42.880 | The typographic power of the web has gotten better,

01:44:47.440 | but it's still not on the level of PDF

01:44:49.320 | and you can't do advanced typography,

01:44:51.560 | but it's gotten really better.

01:44:53.160 | And even then tables were getting better.

01:44:54.920 | If you were using Firefox, that would have been 2004

01:44:57.280 | 'cause it was called Firebird until earlier that year.

01:45:00.240 | - No, yeah, I think it wasn't.

01:45:01.840 | - 2003. - Well, I don't remember.

01:45:03.640 | It was a Firebird, which had tabs.

01:45:06.120 | - We had tabs the whole way.

01:45:07.080 | So it started out as Mozilla/browser in 2002,

01:45:10.440 | became Phoenix.

01:45:12.080 | There's a BIOS that has an embedded version of IE

01:45:14.680 | and they said, "We're called Phoenix Technologies.

01:45:16.360 | "You can't use Phoenix."

01:45:17.600 | And so we said, "Okay, we'll call it Firebird."

01:45:19.160 | And then this Australian-centered

01:45:21.200 | open source database project started really

01:45:24.080 | like in the true Mad Max style, just screaming at us,

01:45:27.120 | saying, "You can't use Firebird."

01:45:28.320 | And I had to sort of be the ambassador.

01:45:30.440 | I said, "Okay, we're gonna rename it."

01:45:31.280 | And they're like, "We don't believe you.

01:45:32.120 | "You shouldn't have used it.

01:45:32.960 | "We hate you."

01:45:33.800 | And then we renamed it to Firefox.

01:45:36.080 | And they're like, "Oh, we love you."

01:45:38.040 | And then I haven't heard of them ever since.

01:45:40.080 | But Firefox was a clever name.

01:45:41.740 | We had to think of something distinctive.

01:45:43.600 | We wanted to keep the fire going.

01:45:45.720 | And it turns out there's a red panda, right,

01:45:48.160 | it's a nickname for.

01:45:49.440 | - So that's the second set of browser wars.

01:45:51.200 | - Second browser war.

01:45:52.040 | - So how did you, how was Firefox born?

01:45:57.000 | How's Mozilla born?

01:45:58.640 | Is there a--

01:45:59.600 | - There's a long story there too.

01:46:01.020 | So Netscape got acquired by AOL,

01:46:02.960 | which as I say was a reasonable happy ending

01:46:05.460 | for a lot of people because Netscape otherwise

01:46:07.600 | was gonna go out of business

01:46:08.440 | because Microsoft was just killing its market.

01:46:10.700 | There was no way to charge for a browser.

01:46:12.480 | Windows came with IE.

01:46:14.760 | IE 4 was pretty good and Netscape 4 wasn't that good.

01:46:17.560 | It took a while to get better.

01:46:19.060 | But the Netscape executive said,

01:46:22.260 | "Let's do an open source escape pod."

01:46:25.640 | And like in Star Wars and New Hope,

01:46:27.120 | the gunner won't shoot it 'cause there's no life forms

01:46:29.240 | on board, right?

01:46:30.160 | It's not a threat.

01:46:31.400 | And so we did Mozilla in 1998 and it looked like

01:46:35.400 | it was going to initially just give the world

01:46:38.480 | an open source browser.

01:46:40.080 | But it's really hard to get people to work

01:46:41.960 | on this sort of hairball that had been hacked up

01:46:43.880 | over by that point four years.

01:46:46.060 | It also couldn't have the crypto module

01:46:49.080 | for secure sockets, so-called,

01:46:50.960 | or now transport layer security.

01:46:52.900 | That was an electronic munition.

01:46:54.320 | We were not allowed to release that

01:46:56.040 | in the full 1024-bit key strength.

01:47:00.400 | And yet people, one of whom I happened to meet previously

01:47:05.000 | at SGI when I went on a sales support engineering trip,

01:47:08.900 | Tim Hudson in Brisbane, Australia,

01:47:11.020 | and Eric A. Young did what became OpenSSL.

01:47:16.020 | It was called SSLEAY after Eric's initials.

01:47:19.440 | And Tim and Eric took their OpenSSL

01:47:23.100 | outside of the purview of the NSA

01:47:25.360 | and the Department of Commerce,

01:47:27.200 | and they stuck it into Mozilla's code.

01:47:28.880 | And that was perhaps the best hack that was done

01:47:31.000 | in the first few months after we open sourced the browser.

01:47:34.460 | We had other problems.

01:47:35.360 | The politics inside Netscape were riven

01:47:37.320 | by these acquisitions.

01:47:38.280 | So the one acquisition that kind of messed up Netscape 4

01:47:41.520 | also wanted to keep doing a proprietary mail

01:47:44.280 | and groupware program, not Jamie Zawinski's mail program

01:47:48.780 | that was in Netscape 2 and 3.

01:47:50.080 | And they held it back from open source.

01:47:51.800 | So we didn't have a mail program, it was just a browser.

01:47:54.320 | We didn't know what AOL would do to us.

01:47:56.720 | Turns out they didn't interfere with us for a long time.

01:47:59.280 | But Netscape wasn't the best steward of Mozilla.

01:48:03.000 | We were operating Mozilla as a pirate ship

01:48:05.320 | without a legal entity.

01:48:06.400 | So most of us worked for Netscape

01:48:08.900 | under a separate organization.

01:48:10.920 | And initially the first engineering manager, Tom Paquin

01:48:15.920 | of Netscape was the Mozilla founding manager.

01:48:19.360 | But he left pretty quickly.

01:48:20.840 | And he left me as the acting manager,

01:48:23.120 | which is more like method acting in my case.

01:48:25.520 | And that was my first management stint.

01:48:28.800 | But then someone who'd written the licenses, Mitchell Baker,

01:48:32.960 | she was a lawyer at Netscape.

01:48:34.120 | She was involved in the open source license decision-making

01:48:38.200 | and the actual writing and construction of those licenses.

01:48:40.400 | That was Mitchell's job.

01:48:41.880 | Netscape public license

01:48:43.000 | and the truly open Mozilla public license.

01:48:45.680 | And there were two because Netscape needed,

01:48:47.400 | because of some encumbered code, needed some special rights,

01:48:49.560 | but that went away over time.

01:48:51.080 | Mitchell was always interested in Mozilla

01:48:52.840 | and she came back from maternity leave.

01:48:54.200 | And she said, "I'll be the manager if you want."

01:48:55.920 | And Jamie and I said, "Sure."

01:48:57.640 | And then Jamie quit.

01:48:58.480 | He quit after a year.

01:48:59.400 | He said, "This didn't work.

01:49:00.840 | I'm sorry."

01:49:01.680 | He acted like it was a total failure

01:49:04.360 | because Mozilla didn't restart the browser market.

01:49:08.080 | But there's no way it could have.

01:49:09.280 | Netscape was still shipping variants of Netscape 4,

01:49:13.760 | which was based on the old code.

01:49:15.720 | Mozilla was trying to re-architect code

01:49:18.400 | to make greenfield for developers.

01:49:19.880 | So it was one of my big goals.

01:49:20.920 | It wasn't a technical goal so much as, again, a social goal.

01:49:24.160 | People wanted a more standard spaced browser.

01:49:26.400 | They wanted less of a hairball that had been hacked on

01:49:29.400 | by ex-grad students starting four years prior.

01:49:32.040 | So we said, "We're gonna make a modular code base.

01:49:35.120 | We're gonna use a variant or an open source version

01:49:37.560 | of Microsoft's component object model,

01:49:40.000 | has reference counting and standardized V tables,

01:49:43.480 | virtual calls and C++.

01:49:45.240 | And we're gonna use JavaScript.

01:49:47.360 | We're gonna have a bridge between those two

01:49:48.720 | so you can script those components

01:49:50.160 | just like Java components.

01:49:51.560 | We're going to make a portable front end

01:49:54.800 | with a markup language for the user interface.

01:49:57.200 | Not tables, not HTML, but custom menus

01:50:00.400 | and dropdowns and toolbars."

01:50:02.880 | And that was called Zool, XML user interface language.

01:50:06.080 | And some real talent on the Netscape side

01:50:07.720 | delivered that, Dave Hyatt, who was instrumental in Zool,

01:50:11.320 | Chris Watterson, Joe Hewitt, Blake Ross.

01:50:15.680 | And Blake was an intern.

01:50:17.800 | He was like a high school aged intern at Netscape.

01:50:20.480 | And at some point we were innovating rapidly

01:50:23.920 | in the Mozilla world and Netscape was still caught up

01:50:26.960 | in this management mess from these acquisitions.

01:50:29.360 | And it wasn't delivering.

01:50:30.440 | And every year they were wondering if AOL was gonna come

01:50:32.800 | and start beheading the executives

01:50:34.280 | 'cause it didn't do anything useful.

01:50:36.080 | And there was this thought,

01:50:36.920 | you should take the Netscape browser engine

01:50:38.920 | and put it in the Windows AOL client,

01:50:40.720 | which was the dial-up client that all the increasingly aging

01:50:43.600 | users of AOL were using.

01:50:45.400 | Never happened.

01:50:46.240 | It would have been too big a change.

01:50:48.120 | So it wasn't clear why AOL bought Netscape,

01:50:49.760 | but as I said, they left it alone,

01:50:50.880 | but Netscape didn't leave Mozilla alone.

01:50:53.280 | And so in 2001, Mitchell called me up and said,

01:50:58.280 | "I'm no longer employed."

01:50:59.800 | And I was like, "What, you quit?"

01:51:00.840 | And no, no, this wasn't my choice.

01:51:02.320 | And there was a layoff, which maybe accidentally

01:51:05.280 | or on purpose got rid of Mitchell.

01:51:06.800 | But the funny thing was we had an open source project.

01:51:08.640 | We had a lot of the engineers on staff on our side

01:51:11.200 | and we had people we'd hired through the Mozilla community

01:51:15.200 | who were top-notch.

01:51:16.280 | They'd risen, they came in high quality,

01:51:18.640 | they knew the code, and they actually were better

01:51:20.440 | than the average or median hire of Netscape.

01:51:23.320 | And so the funny thing was the executive

01:51:26.920 | who thought they'd gotten rid of Mitchell on the layoff,

01:51:29.920 | on the next week's community call around Mozilla

01:51:32.640 | and what to do, there's Mitchell.

01:51:34.240 | And so this showed you can kind of transcend

01:51:36.760 | your boundaries of corporate open source

01:51:40.400 | if you get a project that has enough loyalty,

01:51:42.680 | even among the paid staff.

01:51:43.760 | 'Cause we had outside people contributing.

01:51:45.520 | We had people at Red Hat and a few other places,

01:51:47.920 | but the majority of the hackers were employed by Netscape.

01:51:50.560 | But a lot of them at that point had come from the community

01:51:53.440 | and others got the community and wanted to work with it.

01:51:55.920 | And it was really the weakest engineers at Netscape

01:51:57.840 | who didn't like Mozilla and didn't like the crucible

01:52:01.040 | of competing with the better programmers.

01:52:04.160 | - So if the project is good enough, it will rise,

01:52:06.600 | the Phoenix will rise out of the--

01:52:08.160 | - That's exactly right.

01:52:09.000 | And so we had this Mozilla code base

01:52:10.840 | that was getting better.

01:52:11.680 | In fact, I think at some point in 2002,

01:52:13.480 | when we declared Mozilla 1.0, I engineered a roadmap

01:52:17.320 | that successively through similar sort of six week,

01:52:20.040 | five week releases, like we all do

01:52:21.840 | with browser releases nowadays,

01:52:23.400 | Chrome does and Firefox braved us three weeks.

01:52:26.520 | We got to a point where we said, you know what?

01:52:29.640 | It doesn't suck.

01:52:30.480 | This is like the 1.0 that you want to release

01:52:33.160 | because if you hold it back any longer to polish it,

01:52:35.280 | you're denying others the ability to use it.

01:52:36.920 | It's like Pro Engineer, the mechanical CAD tool,

01:52:39.320 | embedded the code, they embedded the layout engine.

01:52:43.600 | And Mozilla 1.0 was like a Netscape communication suite.

01:52:46.920 | We had at that point gotten male people

01:52:49.520 | to reintegrate mail and news,

01:52:51.120 | and we had an editor for HTML.

01:52:53.560 | And it felt like a 90s suite, suite where.

01:52:56.560 | And it felt kind of bloated.

01:52:58.760 | And the people who were taking that Mozilla open source

01:53:02.080 | and then adding Netscape flavor to it

01:53:04.960 | were not calling the shots right.

01:53:06.480 | And they were also under ALS thumb a little bit

01:53:08.400 | in that they said, well, we should probably put

01:53:10.440 | the ALS instant messenger chicklet on the toolbar.

01:53:13.080 | And we should put the ICQ,

01:53:14.920 | the other messaging system that ALS had acquired.

01:53:17.240 | We should put the ICQ button on the toolbar.

01:53:19.520 | And pretty soon Netscape looked like a bit of a NASCAR

01:53:23.440 | badged version of Mozilla.

01:53:24.960 | And that also made Mozilla more popular.

01:53:27.080 | And yet they had contrived to fire or lay off the leader.

01:53:32.080 | And we'd carried on with an open source structure

01:53:38.440 | where Mozilla was still, you know, Mitchell was calling

01:53:40.800 | sort of management or project level shots.

01:53:44.400 | And I was calling technical shots.

01:53:46.040 | And we had a popular suite, but we thought,

01:53:51.040 | why not make it just a browser?

01:53:52.640 | Because it'll be simpler, it'll do one job well.

01:53:55.480 | And even then we can strip it down by having extensions.

01:53:58.720 | So Dave Hyatt and Blake Ross, the high school aged intern,

01:54:03.640 | did the first version, which was called Mozilla/browser.

01:54:07.200 | It was a small group of us, Ian Hicks and Asa Dotsler,

01:54:09.880 | me and Joe Hewitt and Hyatt and Blake.

01:54:14.280 | And Hyatt was really the senior hacker.

01:54:16.480 | He'd done all these things like amazing cross-platform menus

01:54:20.440 | through the user interface, market language.

01:54:23.440 | And he knew how to do tab browsing.

01:54:24.680 | He'd implemented it natively on Mac OS at the time

01:54:28.200 | in Camino, originally called Camara.

01:54:31.840 | He'd written multiple implementations,

01:54:34.240 | which was a thing programmers should do.

01:54:36.280 | It's like the V8 team did for those missing years

01:54:38.400 | when the rest of the Chrome team's like, where's V8?

01:54:41.000 | In fact, Dave's wife, Rebecca, told me a story

01:54:44.400 | about when they were at UIUC,

01:54:45.800 | they were also University of Illinois grad students.

01:54:49.400 | There was an assignment, it was a programming assignment,

01:54:51.240 | it was supposed to be due at the end of the semester.

01:54:53.240 | And Dave's friend was this, I'm gonna go think

01:54:56.800 | and I'm gonna design and I'm gonna make this platonic,

01:55:00.080 | perfect form of the program.

01:55:01.240 | And then I'm gonna write it at the end when it's due.

01:55:03.560 | And Hyatt just went in there and started hacking.

01:55:05.080 | And he wrote one version, he wrote a second version,

01:55:06.640 | a third version, end of the semester comes around,

01:55:09.240 | the friend's not doing too well.

01:55:11.160 | It wasn't perfect and it wasn't written.

01:55:13.000 | I'm not sure how that story ended for him,

01:55:14.800 | but Dave's version was a fifth iteration, it was great.

01:55:18.080 | And so he'd done that with everything you need

01:55:20.680 | in a tabbed browser.

01:55:21.640 | And this really showed well in Phoenix,

01:55:24.320 | what we called Phoenix and had to rename two more times.

01:55:27.120 | And Blake went to Stanford, he became a Stanford student

01:55:32.240 | and couldn't work on it.

01:55:34.000 | Dave Hyatt went to Apple in 2001.

01:55:36.560 | He was one of the founding Safari team members.

01:55:39.520 | - Interesting, wow.

01:55:40.440 | - But he was still blogging about tabbed browsing.

01:55:43.480 | I think Apple at some point said--

01:55:44.320 | - Did Safari have tabbed browsing?

01:55:46.080 | - Yeah, but it was because of Hyatt.

01:55:48.160 | Hyatt was quite a feather in their cap.

01:55:49.760 | Don Melton, who had been the engineering manager

01:55:53.360 | for Safari from the beginning, had been in that scape also.

01:55:56.480 | And so this is a diaspora of talent.

01:55:58.920 | And yet Hyatt was still kind of writing blog posts

01:56:01.520 | about how to do tabs right.

01:56:03.200 | And at some point Apple said, "Don't blog about that.

01:56:05.240 | "That's our proprietary tab technology."

01:56:07.560 | And it's like, "No, it's not.

01:56:08.480 | "It was an opera and I've refined it."

01:56:10.400 | So we had to replace people.

01:56:12.680 | And we had Ben Goodger, a New Zealander,

01:56:17.120 | we hired at Netscape, and he stepped in

01:56:18.920 | to be the Firefox lead.

01:56:21.240 | And we also had this weird circumstance

01:56:23.320 | where AOL finally did notice that Netscape

01:56:26.880 | was kind of an albatross, that they bought it

01:56:28.960 | for no particular benefit.

01:56:31.040 | And even then the AOL politics were also heinous,

01:56:33.840 | sort of East Coast politics.

01:56:35.320 | I remember taking two trips there

01:56:36.640 | 'cause I was a principal engineer.

01:56:37.760 | And so us principal engineers got trotted out

01:56:39.520 | to do dog and pony shows in Dulles, Virginia.

01:56:42.720 | And the AOL opera management was very East Coast in flavor.

01:56:46.360 | And they were at that time merging with Time Warner,

01:56:49.040 | which did not go well.

01:56:50.480 | So one of these years we went out there

01:56:51.960 | and we were all doing dog and pony shows.

01:56:53.400 | And there were these characters

01:56:55.040 | that were sort of like marketing guys,

01:56:56.320 | one of them was wearing a cravat,

01:56:57.640 | and one was named Reggie.

01:56:59.400 | And they were very you rather than non-you,

01:57:04.400 | or they were like what's Stillman's Metropolitan film,

01:57:08.240 | UHB, urban haute bourgeoisie.

01:57:13.240 | They were haute bourgeoisie, they were funny,

01:57:17.400 | and they were kind of useless and kind of preppy.

01:57:20.080 | And then the next year we went back and I said,

01:57:21.680 | where's Reggie?

01:57:22.520 | And it's like, oh, Reggie's not here anymore

01:57:24.080 | 'cause Time Warner realized that the merger

01:57:26.240 | wasn't in their interest either.

01:57:27.440 | And then the sort of knives came out.

01:57:29.080 | And these mergers rarely work, right?

01:57:31.760 | This is very difficult.

01:57:32.600 | You get these giant companies

01:57:33.760 | and they think there's gonna be synergy.

01:57:35.240 | That was the late '90s with Watchword.

01:57:37.280 | And there wasn't synergy with AOL buying Netscape,

01:57:39.280 | and there wasn't synergy with Time Warner and AOL.

01:57:41.200 | - But did AOL ever really work?

01:57:42.960 | Was it ever really cool?

01:57:44.240 | Like the same kind of fire and excitement

01:57:46.440 | that Firefox eventually created,

01:57:48.720 | was that ever there in AOL?

01:57:51.080 | - AOL was the right time to do a dial-up service

01:57:55.000 | that got distribution by basically

01:57:56.640 | leaflet-bombing compact disks on the country.

01:58:00.680 | And they beat out CompuServe and the other ones, Prodigy,

01:58:05.520 | and then the web happened.

01:58:06.920 | And so you had almost like this isolated continent,

01:58:10.920 | like some of the evolutionary biologists I follow

01:58:14.160 | make fun of the funny large marsupial mammals of Australia,

01:58:19.160 | how silly they are.

01:58:21.560 | - So AOL's like Australia.

01:58:23.080 | (Lex laughing)

01:58:23.920 | - And you saw it over time because they kept aging

01:58:26.420 | and they were using AOL to get online

01:58:28.200 | and they couldn't really use a web browser.

01:58:30.480 | And it became sort of a valued cohort

01:58:33.440 | 'cause they still have relatively high socioeconomic status

01:58:36.320 | and they have grandchildren, but it's going away.

01:58:38.480 | It's dying at some point.

01:58:39.520 | - Towards the end of the aughts, that decade,

01:58:42.160 | and then to the decade 2010 plus,

01:58:46.040 | that Firefox became this incredible,

01:58:48.680 | I forget when Chrome came out, but--

01:58:50.280 | - 2008, September.

01:58:51.120 | - 2008, but Firefox was the sexy, cool thing

01:58:55.120 | that represented a lot of the cutting edge technologies

01:58:57.540 | and all that kind of stuff.

01:58:58.380 | - Web 2, it was amazing.

01:59:00.180 | Tim O'Reilly and John Battelle did the first Web 2

01:59:02.900 | conference, which eventually became huge

01:59:04.460 | and they split it, but that was in 2004.

01:59:06.700 | That's right when Firefox was out.

01:59:08.580 | Craigslist was huge, it was killing classified revenue

01:59:11.700 | for newspapers, but there was just this ferment.

01:59:14.460 | People starting--

01:59:15.300 | - Wikipedia along there somewhere?

01:59:16.700 | - Gmail was already done and it was an impressive web mail.

01:59:19.860 | There were others before it like Hotmail,

01:59:21.180 | but Gmail was really impressive from Google.

01:59:22.940 | And Google Maps, people started seeing what could be done.

01:59:26.040 | They thought, how can you drag the map around

01:59:28.120 | and how does that work?

01:59:29.720 | And it was all JavaScript and images.

01:59:31.520 | - Gmail was 2003, four?

01:59:34.520 | - Yeah, I think it actually started quite early.

01:59:36.240 | It might have been 2002 or three, but by the time

01:59:38.460 | we started dealing with Google and Firefox

01:59:40.120 | to get the search deal, which was the main revenue source

01:59:42.520 | for Mozilla, and still is, 2004, early,

01:59:46.600 | Sergey Brin's, one of his trusted engineer guys,

01:59:49.760 | Fritz Schneider, made contact with me at Mozilla

01:59:51.960 | and we started talking and we realized search

01:59:55.420 | and browser need each other.

01:59:57.020 | And this is deeply true, right?

01:59:58.500 | This is still true.

01:59:59.820 | This is why a lot of the search engines

02:00:01.180 | have their own browsers.

02:00:02.300 | - Yeah, so in case people don't know,

02:00:03.700 | the main revenue source for the browser

02:00:05.620 | is the default search engine, which is kind of incredible

02:00:08.620 | to think about, that that is a revenue source.

02:00:11.500 | It's a little bit sad.

02:00:12.620 | - Yeah, it leads to this capture or kill effect

02:00:14.620 | where you have the search engine own its own browser

02:00:17.060 | and other browsers may struggle to get distribution,

02:00:21.300 | we talked about it earlier.

02:00:22.640 | - So where, and you said you've figured out

02:00:26.820 | that Google's working on its own browser at some point.

02:00:29.780 | - 2006, yeah.

02:00:30.760 | - 2006, so would you say Firefox versus,

02:00:33.260 | was Internet Explorer part of the war here

02:00:35.420 | or was it Firefox versus Chrome?

02:00:37.320 | - So Firefox didn't quite cause Microsoft

02:00:39.560 | to reconvene, they did do IE7, and I remember

02:00:43.760 | being on a plane back from the standards meeting,

02:00:46.240 | JavaScript standards meeting from Seattle, from Redmond,

02:00:49.000 | and there was some Microsoft guy in front of me,

02:00:52.120 | turns out my wife knew him from her past life

02:00:54.920 | before we married, and he was just this bearded big guy

02:00:58.200 | and he was like, we should have just killed Firefox

02:01:00.920 | in the cradle, all we needed to do was add popup blocking

02:01:03.520 | and tabs and we could have made Internet Explorer

02:01:05.560 | kill Firefox, and it's like, shoulda, coulda, woulda, pal,

02:01:07.840 | and I was right behind him overhearing this.

02:01:10.320 | But they didn't, they were slow and IE7 wasn't that great,

02:01:13.520 | and what really got them started, I think, was Chrome.

02:01:18.440 | And I talked to Larry Page in 2005, I think I said,

02:01:22.800 | we're talking about the Firefox relationship,

02:01:24.500 | but he was also saying, what about WebKit?

02:01:26.120 | This was Apple's version of the old KHTML engine

02:01:30.480 | from Linux, the KDE side of Linux, that was used

02:01:34.640 | in the Conqueror browser, also with Ks,

02:01:36.880 | that Apple had forked, and in 2005 was when Apple's

02:01:41.240 | principals, including Dave Hyatt, Maciej Stachowiak,

02:01:43.400 | some of my friends who are still there, said,

02:01:45.200 | we must stop patch bombing this poor KHTML project,

02:01:48.560 | we should make a proper Mozilla-like organization,

02:01:50.920 | WebKit.org, now it wasn't a separate non-profit

02:01:53.340 | or anything, it was still Apple, it was Apple-controlled,

02:01:55.240 | but they made their fork first class and they made it

02:01:58.960 | be something that they all worked in and lived in,

02:02:01.440 | and that was before Chrome, and then Chrome,

02:02:04.080 | Larry Page said, what about WebKit?

02:02:05.680 | I said, yeah, it's nice, I have friends who work on it,

02:02:08.000 | you might use that if you do your own browser,

02:02:09.560 | why don't you do your own browser?

02:02:10.480 | Don't worry about Firefox, you should do your own browser,

02:02:12.920 | you can have your own opinion of how it should work,

02:02:15.280 | and sure enough, they did, so by 2006, we knew

02:02:18.400 | they'd been working on it, some of my friends

02:02:20.160 | who'd been at Netscape did the original demo,

02:02:22.440 | and the demo wasn't what you thought,

02:02:24.020 | it didn't have the fast JavaScript yet,

02:02:25.360 | that was still off in Denmark on a farm.

02:02:28.520 | - Did it have tabs?

02:02:30.520 | - It had tabs, 'cause all browsers had tabs at this point,

02:02:32.800 | and it had this software fault isolation I mentioned,

02:02:37.480 | it was through process isolation, so in theory,

02:02:41.120 | each tab has its own operating system process,

02:02:43.600 | and so what's gonna take your tab down?

02:02:45.640 | Well, WebKit has bugs that can crash it,

02:02:48.440 | but Flash was still big then, all the restaurant sites,

02:02:50.840 | remember, and Flash crashed a lot, so the demo

02:02:54.320 | that I heard about, my friends at Netscape,

02:02:56.520 | as a lot of people did, inside Google was the sad tab,

02:03:00.500 | they showed an early version of Chrome,

02:03:02.160 | which is just this bare bones tab browser,

02:03:04.240 | they loaded a site with a known Flash vuln,

02:03:06.720 | and then suddenly, Flash crashes, and everyone expected

02:03:09.240 | the whole browser to go down, but instead,

02:03:11.320 | you got this little sad face in the tab,

02:03:13.120 | and you could reload it, and there it is again,

02:03:14.800 | so this was an improvement, it was a real move for security,

02:03:18.760 | it was based on a company they acquired called Green Border,

02:03:22.720 | they had some really big brains, like Olfar Erlingson,

02:03:25.080 | I think was involved, and they had done some exotic

02:03:28.160 | security stuff, but they ended up simplifying it

02:03:30.080 | to this process isolation, and it was good,

02:03:34.980 | and Firefox didn't have it at the time,

02:03:36.760 | so we were still struggling with security bugs.

02:03:39.700 | So we knew Chrome was coming, but it took two more years

02:03:43.920 | to come out, and we were still getting the Google search

02:03:47.960 | revenue, and we were still making Google the default engine,

02:03:51.840 | and Firefox was still growing, Firefox grew,

02:03:54.000 | I think until 2011, that was when it peaked,

02:03:56.860 | and as it started falling, it was because of Chrome.

02:03:59.840 | Chrome came out in 2008, and it was a comic book

02:04:02.880 | that leaked accidentally, that showed some of the people

02:04:04.800 | who worked on it, Lars Bach was in there, and so on.

02:04:07.240 | It was kind of soft launch, because they didn't market

02:04:10.040 | it heavily, they didn't push distribution,

02:04:12.560 | but Google had reason to worry about distribution,

02:04:14.800 | because Microsoft was doing a search engine,

02:04:18.760 | Bing, since 2007, in fact, when they came out with Bing,

02:04:22.000 | Google was worried that Microsoft would just brute force

02:04:25.200 | switch the default browser in everyone's Internet Explorer,

02:04:28.840 | or even Firefox on Windows, to Bing from Google,

02:04:32.520 | and Microsoft wasn't, I think, ready to dare

02:04:35.760 | the antitrust cops that way, even though

02:04:37.680 | they'd gone to sleep, and I don't think

02:04:40.360 | Bing was ready, either, but just in case it happened,

02:04:44.000 | Sundar Pichai, who rose very well based on this work,

02:04:47.760 | was sort of in charge of getting distribution deals,

02:04:50.600 | and he got Google Toolbar, and Google Desktop Search

02:04:54.920 | distribution, and if you remember those pieces of software,

02:04:56.800 | those were like desktop extensions, toolbars,

02:05:00.680 | or operating system extensions for doing desktop search

02:05:03.080 | in your local files, kind of like--

02:05:03.920 | - Oh, that's right. - Mac OS Spotlight.

02:05:05.880 | - Sadly died out. - It all died,

02:05:07.800 | and there were some features that we still missed

02:05:09.200 | that didn't make it into Chrome,

02:05:10.520 | but Sundar got OEMs to bundle those,

02:05:13.960 | and then he got enough of those deals

02:05:15.440 | that by 2007 or '08, Google felt, well,

02:05:18.400 | if Microsoft does the worst and tries to force Bing,

02:05:21.800 | we can reach in and reset it with that point of presence.

02:05:25.400 | So that was good for Sundar's career,

02:05:27.240 | and it was good for Google, but it never came to pass

02:05:29.160 | that they had to defend.

02:05:30.720 | Microsoft was still slow, and by the time they saw

02:05:34.240 | Chrome come out, then they did what would have been IE9,

02:05:37.360 | and then they said, we're gonna have a fast

02:05:38.720 | JavaScript engine to Chakra, Chakra Core,

02:05:41.360 | and they did okay.

02:05:43.160 | They were another process-isolated,

02:05:46.240 | fast JavaScript browser, tab browser.

02:05:48.960 | - So it sounds like there's a deep,

02:05:51.320 | fundamental coupling of search engine and browser

02:05:53.640 | that's mixing this whole thing up,

02:05:55.760 | and obviously Firefox doesn't have a search engine

02:05:59.960 | that's like, I mean, you're partnering with somebody

02:06:03.680 | with a search engine, with Yahoo or with Google or so on.

02:06:08.680 | - They tried Yahoo.

02:06:09.600 | That was unfortunate because I think,

02:06:11.400 | even though Marissa Mayer talked about it,

02:06:14.600 | she never pulled it off.

02:06:15.800 | They never restored the search team

02:06:18.320 | that had been laid off.

02:06:19.560 | I believe Carol Bartz was running Yahoo

02:06:21.440 | when Carol said, I've gotta get rid of

02:06:23.320 | one of the three expensive things.

02:06:24.520 | I'm gonna get rid of search.

02:06:25.920 | And those researchers went to Google and Microsoft,

02:06:29.760 | and there was no way to put Yahoo search back together.

02:06:32.480 | So when Firefox tried switching all their users

02:06:36.000 | who'd stuck with a default from Google to Yahoo,

02:06:38.840 | like mid-December 2014, a bunch of users said,

02:06:41.880 | what just happened to my Firefox?

02:06:44.040 | And others didn't notice right away,

02:06:45.280 | but over time they did, and so over the next year,

02:06:48.280 | the traffic just went away for Yahoo,

02:06:51.320 | and yet they were obliged, I understand it,

02:06:54.160 | I don't have inside knowledge, but this is leaked out,

02:06:56.640 | and Danny Sullivan's written about it,

02:06:58.320 | search engine, and I think the deal

02:06:59.920 | was like fixed payments to Mozilla.

02:07:01.480 | So Mozilla was getting a bunch of money for traffic

02:07:03.480 | that wasn't staying because users were resetting

02:07:05.880 | their default, and this shows how defaults are important,

02:07:09.200 | but they have to be good enough

02:07:10.520 | that the user doesn't override them.

02:07:12.560 | And a lot of the commercial value in popular apps

02:07:16.600 | is what are the default settings?

02:07:18.040 | What is the default search?

02:07:20.120 | - But oftentimes there's something, just like you said,

02:07:21.840 | I mean, if there's something compelling,

02:07:25.040 | that also can beat out the default,

02:07:27.240 | like tab browsing and so on.

02:07:29.040 | And that's where, I mean, we'll talk about Brave browser.

02:07:32.320 | It feels like now we're in this third stage

02:07:36.200 | where there's Chrome, Firefox, Edge, I guess it's called,

02:07:41.200 | and Brave, and these all seem like really exciting,

02:07:48.240 | I don't know, innovative browsers.

02:07:50.560 | They're all kind of copying off of each other,

02:07:52.240 | picking up the good stuff.

02:07:53.760 | - There's evolution again,

02:07:54.640 | especially on tracking protection.

02:07:56.240 | So privacy is this sort of global wave that's rising.

02:08:00.240 | I like to call it a wave

02:08:03.360 | because it's a large, somewhat chaotic structure.

02:08:07.040 | It's not a unitary good.

02:08:08.600 | You can't say I'm buying privacy for $3.

02:08:11.760 | I'm paying $3 for privacy.

02:08:12.720 | Some people think a VPN does this

02:08:14.280 | and are disappointed when it fails them,

02:08:16.000 | but often people use VPNs for region unlocking video

02:08:19.720 | or getting the US Netflix catalog.

02:08:21.400 | - Exactly.

02:08:22.840 | - But privacy is not a unitary good.

02:08:24.720 | It's complex and people are understanding it

02:08:26.880 | only over time and as they get burned,

02:08:28.920 | but there's a genie

02:08:30.680 | that's not going back in the bottle there.

02:08:31.960 | People are fed up.

02:08:33.360 | Apple has responded to this.

02:08:34.680 | Apple was always making Safari, I think,

02:08:36.600 | more of a privacy-branded browser from the very beginning.

02:08:40.320 | I think this was probably Steve Jobs.

02:08:42.280 | Safari had private windows, private tabs,

02:08:45.280 | before Firefox did.

02:08:48.240 | And these are only private in the sense

02:08:50.080 | that they don't leave local traces

02:08:53.040 | if you don't want them to.

02:08:54.000 | Turns out Safari does keep them around between shutdown,

02:08:57.200 | but the canonical model is no local traces

02:08:59.880 | after you close the private window.

02:09:01.800 | No leftover traces that you went to some site

02:09:04.960 | that you were embarrassed by

02:09:06.160 | or bought a gift for somebody you wanted to keep secret.

02:09:08.960 | - But there's still some level of tracking.

02:09:11.800 | - There's network tracking.

02:09:14.880 | Network privacy is not guaranteed at all

02:09:16.920 | because you're using the same internet and ISP

02:09:19.000 | as a public window, a non-private window.

02:09:21.360 | But Safari had that early on.

02:09:22.920 | They also had a cookie-blocking policy

02:09:26.040 | that might take a little explaining.

02:09:28.000 | If you know what a cookie is,

02:09:29.800 | it's a little bit of storage in the browser

02:09:31.320 | indexed by the name of the site.

02:09:33.400 | And it's really only the main name of the site,

02:09:35.040 | like bofa.com or something like npr.org.

02:09:40.040 | Every site can store some information in a cookie.

02:09:46.000 | Every time it's contacted by the browser,

02:09:48.640 | the previous version is sent back.

02:09:50.880 | And in the response from the server, the cookie's updated.

02:09:54.080 | So it's this little bit of storage in the browser

02:09:56.600 | that the site can keep updating,

02:09:58.200 | and it can store an encrypted version

02:10:01.480 | of your login credentials with a timestamp

02:10:03.800 | so you can stay logged in

02:10:05.960 | without having to retype your password

02:10:07.720 | every time you navigate,

02:10:08.840 | which is how it would be if you didn't have cookies.

02:10:11.240 | The web protocols, especially in the '90s,

02:10:14.440 | are so-called stateless protocols.

02:10:16.000 | So you go to your bank, you log in,

02:10:18.760 | you go from your login confirmed page to your account view.

02:10:22.480 | If you didn't have a cookie, you'd be logging in again.

02:10:25.000 | - Every time you type in the password.

02:10:26.560 | - So that was a great thing about cookies.

02:10:27.880 | Lou Montuoli did it in a hurry in 1994

02:10:29.840 | before I joined Escape,

02:10:30.880 | and he did it for really holding that kind of credential.

02:10:34.840 | But even then, there was the image element

02:10:37.520 | embedded in the page,

02:10:38.560 | and the image gets fetched possibly from a different server,

02:10:42.000 | and that request carries the last cookie,

02:10:44.240 | which could be empty at first,

02:10:45.760 | and the response carries the updated cookie.

02:10:48.000 | So just by having images and cookies,

02:10:49.760 | you got tracking because that image server

02:10:51.440 | can be serving a little one-by-one pixel,

02:10:54.240 | and they still use the word pixel in ad tech,

02:10:56.520 | and that pixel can be served from the same server

02:10:59.360 | embedded differently with different URL spellings

02:11:02.160 | in the New York Times and ESPN.

02:11:03.760 | And as you go from one to the other,

02:11:05.880 | the image server can say,

02:11:07.000 | I haven't got a cookie for you, it's empty initially.

02:11:09.000 | I'm gonna assign you user number 1234,

02:11:11.240 | I'm gonna put a database entry in,

02:11:13.000 | and I see, by the way, I always fetch

02:11:14.480 | the name of the path part of the URL

02:11:15.960 | that I was in the New York Times.

02:11:17.440 | So you're a New York Times reader.

02:11:18.920 | And then you hit ESPN, same thing,

02:11:20.960 | and the database gets updated,

02:11:22.240 | and the number, user 1234 indexes in the database

02:11:25.520 | to a profile of you, you've been tracked.

02:11:28.080 | This was not intended,

02:11:29.360 | and it was too late to undo by the time I got to Netscape.

02:11:32.680 | I think Lou wanted to do Twinkies, he called them,

02:11:35.040 | and he was trying to solve several problems.

02:11:37.320 | He wanted them to be bigger

02:11:38.320 | 'cause initially cookies had a short size limit.

02:11:40.520 | I think he wanted to solve the third-party problem,

02:11:42.280 | but Tom Paquin, the engineering manager said,

02:11:44.960 | "Nope, no Twinkies, just cookies.

02:11:47.080 | "We're done, you're done, son."

02:11:49.040 | And that's how a lot of that stuff was.

02:11:51.480 | That's how JavaScript got frozen like a flying amber

02:11:55.400 | in some ways with that sloppy equality operator

02:11:57.320 | that I made 'cause of the early adopters.

02:11:58.960 | And the cookie got stuck with this tracking hazard,

02:12:01.840 | and then, because JavaScripts can be like images,

02:12:04.240 | they're embedded in the page.

02:12:05.880 | By the time of Netscape 3, I made that work.

02:12:08.640 | You can get a request with the last cookie value,

02:12:10.760 | and the response updates it.

02:12:11.880 | That's a tracking mechanism,

02:12:12.920 | and that's why you don't even need images to track.

02:12:14.880 | Now you just use scripts.

02:12:16.440 | So this whole tracking economy evolved,

02:12:20.000 | and it depended on these accidents of the '90s,

02:12:23.840 | these unintended consequences.

02:12:25.320 | - Well, it created some of the richest companies

02:12:27.200 | in the world, right?

02:12:28.040 | I mean, it's the social media--

02:12:28.880 | - All I got was T-shirts.

02:12:29.960 | (laughing)

02:12:31.160 | - All I got is this crappy T-shirt.

02:12:33.160 | Yeah, I mean, so that's the fundamental problem

02:12:38.160 | the world is facing now.

02:12:39.520 | They're looking at what social media has created,

02:12:42.040 | and they're looking at,

02:12:44.200 | and like a world is looking at itself in the mirror,

02:12:47.040 | and seeing that privacy is actually something,

02:12:51.000 | it's supposed to be like a nice thing to have.

02:12:53.520 | It's something that is, actually should be fundamental

02:12:57.280 | to the way we interact with the world,

02:12:59.240 | as part of our tooling,

02:13:00.600 | and that's where the Brave browser comes in.

02:13:03.520 | And I suppose others as well are playing with this idea,

02:13:06.040 | but Brave is at the forefront of that.

02:13:07.640 | So maybe can you describe what Brave is,

02:13:11.000 | and what are its key principles,

02:13:13.400 | and what's broken, and what is it Brave trying to fix?

02:13:16.120 | - So when I realized that these accidents,

02:13:18.400 | like the third-party cookie,

02:13:19.960 | the image or script that's tracking you,

02:13:21.840 | or the JavaScripts that can do it invisibly now,

02:13:26.040 | that all this stuff wasn't intended,

02:13:28.680 | and that Firefox had supported extensions

02:13:31.640 | that block some of these things,

02:13:32.640 | I thought, probably we should have browsers

02:13:34.720 | just block some of these things by default.

02:13:37.440 | These were not intended, and they're now unsafe.

02:13:39.760 | They're tracking you, there could be data breaches,

02:13:41.720 | malware distribution, bullying, and PsyOps,

02:13:46.720 | and other attacks on people.

02:13:48.420 | Block that stuff, block that JavaScript.

02:13:51.800 | I'm Dr. Frankenstein, I've got to deal with a monster here.

02:13:54.680 | But obviously you go to Gmail,

02:13:56.520 | there's so much a script there

02:13:57.560 | to make that amazing web client,

02:13:59.320 | that's okay, that's first-party JavaScript.

02:14:02.160 | So how do you tell the first from the third party?

02:14:04.000 | And it's not easy, it's not a matter of just

02:14:06.040 | what's embedded from a different server,

02:14:07.920 | because a lot of publishers use benign scripts

02:14:11.560 | from unrelated domains, or apparently unrelated domains.

02:14:14.360 | So you end up having to develop a sort of

02:14:16.320 | human and machine learning practice around blocking.

02:14:19.700 | And at Brave, we did that from the start,

02:14:22.320 | and built a research team to help drive it and automate it.

02:14:25.720 | We realized that protecting people needed machine learning,

02:14:28.920 | and around 2017, in spring,

02:14:30.400 | I talked to my friends at Apple about this too,

02:14:32.680 | and they were also doing what they call

02:14:34.560 | intelligent tracking prevention,

02:14:36.640 | which uses local machine learning in the browser.

02:14:39.320 | And the funny thing is, you know,

02:14:41.600 | great minds think alike,

02:14:42.920 | they were taking their third-party cookie blocker

02:14:44.800 | that was in Safari from the old days,

02:14:46.480 | and making it not have a big loophole.

02:14:48.520 | Because what they did was, in 2003,

02:14:51.160 | when Safari came out, they said,

02:14:52.000 | we're gonna block cookies that are from

02:14:55.220 | those third-party embedded elements,

02:14:57.240 | where you've never visited that site before.

02:14:59.160 | So I'm gonna pick an ad company that got sold to AT&T,

02:15:02.360 | so I'm not picking on anybody unfairly,

02:15:04.000 | appnexus.com.

02:15:05.120 | Have you ever been to appnexus.com?

02:15:06.640 | - Nope. - I've never been there.

02:15:07.800 | But I guarantee you, 10 years ago,

02:15:09.020 | you probably had, if you were using Firefox,

02:15:11.180 | you had a cookie, third-party cookie,

02:15:13.040 | 'cause you were being tracked by them.

02:15:14.960 | And they were using that cookie

02:15:15.920 | to build up a profile of you.

02:15:17.560 | In Safari, as long as the user never went to AppNexus,

02:15:20.720 | that cookie would not be set.

02:15:22.360 | And that was a real move for privacy early on,

02:15:25.600 | when jobs were still around, in Safari.

02:15:27.660 | But it had this loophole that if you do go to AppNexus,

02:15:30.620 | then why, it's okay to be a third-party cookie.

02:15:33.080 | So AppNexus did something very naughty.

02:15:34.920 | They took their ad partners,

02:15:36.840 | that put the actual ad you click on,

02:15:39.000 | and they said, "Hey, add a little script

02:15:40.240 | "so that when somebody clicks on the ad,

02:15:42.540 | "before it goes to your landing page,

02:15:44.420 | "redirect to AppNexus,

02:15:45.680 | "and we'll redirect to the landing page."

02:15:47.160 | And by doing that, they set a first-party cookie,

02:15:49.120 | and they got whitelisted.

02:15:49.960 | So it was a loophole they exploited.

02:15:51.640 | Intelligent tracking prevention in Safari

02:15:53.560 | was sophisticated enough to counteract this,

02:15:56.600 | and it did other things,

02:15:57.440 | and it's evolved since they did it.

02:15:59.280 | And we've evolved brave, too.

02:16:00.680 | And so when I say machine and human learning,

02:16:02.520 | there's a real set of techniques here.

02:16:05.560 | They have to fight--

02:16:06.400 | - This is a fascinating problem, actually.

02:16:07.560 | - Fingerprinting, right?

02:16:08.400 | Anytime you have a little bit of storage in the browser

02:16:10.680 | associated with a website,

02:16:13.540 | if the bad guy can get 32 websites,

02:16:15.520 | each one has a bit of storage that's 32 bits.

02:16:17.600 | You can turn the bit on or off.

02:16:19.800 | You can make four billion numbers.

02:16:21.240 | You can make an identifier.

02:16:22.240 | It's called a super cookie sometimes.

02:16:24.600 | There are weaker ways that are statistical.

02:16:28.840 | They're called fingerprinting.

02:16:29.760 | You have to block all of them,

02:16:30.740 | and you have to not only automate,

02:16:32.800 | you want to work in the web standards body

02:16:34.920 | to put privacy in by default, by design,

02:16:38.440 | from the get-go, not add it as an afterthought

02:16:40.480 | or go hog-wild with new web APIs

02:16:43.160 | that add a bunch more local storage

02:16:45.240 | or fingerprint surface area.

02:16:47.480 | And that's been a struggle, too,

02:16:48.640 | because guess who's the new Microsoft

02:16:50.880 | in the standards body?

02:16:51.720 | It's Google.

02:16:52.660 | And they're not in favor of privacy first.

02:16:55.880 | They wanna do privacy their way,

02:16:58.720 | only under, I would say, market pressure.

02:17:01.140 | But with Apple and with Brave leading the way,

02:17:03.620 | we block third-party cookies almost without exception.

02:17:06.260 | So we just block them,

02:17:07.900 | and that gives us a very strong privacy benefit,

02:17:11.500 | but it also means some sites just don't work right.

02:17:13.700 | Embedded YouTube videos might not work right.

02:17:15.460 | So we're adapting in a similar way to Apple's done with ITP

02:17:19.580 | to make third-party cookies blocked,

02:17:22.700 | but to sort of simulate what looks like

02:17:26.420 | a working third-party cookie for the site.

02:17:28.840 | It essentially tries to partition each site

02:17:32.320 | and its third parties into its own sort of cookie jar.

02:17:34.840 | - Got it.

02:17:35.680 | And so, like you said,

02:17:37.040 | is this both a human fine-tuning issue

02:17:41.080 | and a machine learning problem?

02:17:42.600 | - Yes, and as the humans learn,

02:17:44.520 | then they train the machine learning.

02:17:46.400 | - But maybe Google aside, or including Google,

02:17:50.560 | there's millions of dollars, if not billions of dollars,

02:17:54.120 | to be made from fighting the ways of Brave.

02:17:57.860 | - That's right, and it's been an interesting change

02:18:00.660 | from when we started in 2015.

02:18:01.980 | When we started, ad-blocking extensions,

02:18:04.460 | AdBlock Plus was one of the big ones

02:18:06.180 | that started on Firefox in 2006, I believe,

02:18:08.660 | had gotten to a certain level of use around the world,

02:18:11.020 | and browsers like UCWeb, UC Browser in Asia,

02:18:14.340 | had some amount of ad-blocking built in and on by default.

02:18:17.260 | So PageFare was a startup,

02:18:19.340 | and they measured ad-blocking adoption,

02:18:23.540 | and they tried to say, "Hey, publishers,

02:18:25.440 | "30% of the visitors to Pitchfork or Wire,

02:18:30.140 | "to Conda NAS properties, are using ad-blockers.

02:18:32.820 | "If we can somehow convince them

02:18:34.300 | "to lower their ad-blocking for your site,

02:18:37.280 | "that could be like a 43% lift, right?"

02:18:39.460 | And 3/7ths, well, that's easier said than done,

02:18:44.300 | and PageFare and others, SourcePoint, many others,

02:18:46.620 | tried to either smuggle ads through

02:18:48.100 | or cajole the user into letting ads appear,

02:18:52.300 | and it didn't really work.

02:18:53.140 | Meanwhile, the ad-blocking adoption

02:18:54.460 | has just continued.

02:18:55.420 | Intelligent Tracking Prevention, Safari in 2017,

02:18:59.340 | Brave from 2016 on with very strong cookie-blocking

02:19:03.140 | and other protections, and this is not going away.

02:19:06.660 | The publishers used to rage against it.

02:19:08.920 | We would try to say, "We can help you.

02:19:10.380 | "You're dealing with users

02:19:11.880 | "who are already blocking all your ads.

02:19:14.040 | "We can try to put back some economics

02:19:16.720 | "that help the user and you,"

02:19:18.460 | that led to the basic attention token

02:19:20.020 | that we started with Bitcoin.

02:19:21.400 | "We can be your friend.

02:19:22.380 | "Don't just fingerprint us as an ad-blocker

02:19:24.900 | "and treat us as an enemy."

02:19:26.500 | But in 2015 or '16, it was like,

02:19:28.560 | "Nah, you're an ad-blocker.

02:19:29.460 | "Get out of here.

02:19:30.300 | "I hate you."

02:19:31.140 | And by 2017 or '18, it's like something's happening.

02:19:33.740 | The ad-blocking's not stopping,

02:19:34.940 | and we're all getting sort of pulled

02:19:37.020 | on the Google's plantation through AMP, A-M-P,

02:19:41.580 | or we're getting killed by the Google ad system we use

02:19:44.540 | because it's taking all the revenue,

02:19:46.580 | or it's permitting, or some other vendors we use

02:19:49.100 | are permitting ad fraud.

02:19:50.100 | And so a fake New York Times is getting paid

02:19:52.620 | by the marketer running an ad that a bot clicks on,

02:19:57.100 | and the real New York Times

02:19:58.220 | that's supposed to get the ad doesn't get it.

02:20:00.500 | - And there's something really broken

02:20:01.540 | about that kind of system.

02:20:02.380 | - And that fraud is mediated through Google's ad exchange,

02:20:06.660 | which is the biggest of them all,

02:20:07.900 | and Google takes a fee.

02:20:09.660 | There's a flip side of that,

02:20:10.700 | which is malware distribution, malvertising,

02:20:12.660 | where fake advertisers put malware payloads

02:20:16.300 | in exploit kit loaders in JavaScript,

02:20:19.340 | and they smuggle them in ads onto real publisher pages.

02:20:22.500 | The ad exchange takes the fee.

02:20:24.500 | Now, I'm not a lawyer.

02:20:25.620 | I'm not gonna say this is a RICO predicate,

02:20:27.140 | but why is the ad exchange facilitating fraud

02:20:30.260 | and malware distribution and taking a fee?

02:20:32.100 | It's not right.

02:20:32.940 | - As opposed to just fighting,

02:20:33.980 | this is the really interesting thing about Brave,

02:20:35.700 | is as opposed to just fighting

02:20:37.700 | and then being treated like an ad blocker,

02:20:39.980 | you're providing an alternate.

02:20:42.060 | There's a philosophical idea here

02:20:45.460 | that might change the nature of the internet

02:20:47.500 | with a basic attention token.

02:20:49.220 | - Yes.

02:20:50.060 | - Maybe what is basic attention token, BAT,

02:20:54.060 | and how does it work?

02:20:55.300 | - Okay, I'll tell the story first by saying how I came to it.

02:20:58.180 | I realized for a long time at Firefox,

02:21:00.820 | we were dependent on this Google search deal,

02:21:02.940 | and I thought, now that Chrome's out,

02:21:05.700 | maybe that's gonna go away.

02:21:07.100 | At some point, Google will say,

02:21:10.740 | Firefox, like old yeller, you saved me from the rabid beast.

02:21:14.620 | Now I have to shoot you in the head.

02:21:16.340 | Done your job, sad but true, goodbye.

02:21:19.660 | And what can we do?

02:21:21.780 | And I think Mozilla doesn't know what to do.

02:21:24.140 | This is something that I couldn't solve there,

02:21:25.900 | and I don't think they can solve.

02:21:27.220 | But I thought, why is the browser

02:21:29.460 | the sort of passive servant of these big tech companies?

02:21:32.300 | Why is it a blind runtime for ad tech JavaScripts,

02:21:36.460 | including from Google?

02:21:38.100 | Why doesn't it block some?

02:21:39.220 | And if it blocks some, why can't it reconnect users,

02:21:43.940 | readers, fans, with publishers, creators, websites?

02:21:47.480 | Why can't it help people make direct payments

02:21:51.220 | or even possibly get an ad revenue share

02:21:53.980 | for private ads that are placed in the browser?

02:21:56.300 | The ads are all placed in the browser.

02:21:57.580 | Some people have this sort of model

02:21:58.820 | that the server's painting the ad into some flash,

02:22:02.360 | combined package or into some giant image,

02:22:06.500 | and then it all gets sent down.

02:22:07.620 | That's not how it works.

02:22:08.600 | All the ads you see on the web are placed in your browser

02:22:11.340 | by it calling out to various ad tech partners,

02:22:13.980 | and Google's among them.

02:22:15.620 | And so if you block those scripts,

02:22:17.560 | you break the advertising flow of money

02:22:22.420 | from the brands and their agencies to the publishers.

02:22:26.300 | And if you want to reconnect it directly with the user,

02:22:29.460 | you have limited choices.

02:22:30.420 | The user generally isn't gonna sign up

02:22:31.940 | with a ACH bank connection or a credit card.

02:22:34.980 | The publisher isn't gonna sign up the user,

02:22:37.260 | except as a subscriber,

02:22:38.360 | and then they're gonna overcharge you

02:22:39.980 | because they want you to cross subsidize all the content

02:22:42.060 | and buy more than you read and all that stuff.

02:22:43.660 | And how many, you know, people are doing great

02:22:46.500 | who are big names like New York Times and The Washington Post

02:22:48.740 | but how many subscriptions are you as a user gonna pay for?

02:22:51.820 | This is why startups like Tony Hale's Scroll

02:22:54.420 | are trying to do a portable subscription system.

02:22:57.180 | - By the way, just on a small tangent there,

02:22:58.860 | even the New York Times is really annoying

02:23:00.900 | how difficult it is just to subscribe.

02:23:03.340 | There's way too many clicks.

02:23:04.660 | - They don't make it easy.

02:23:05.480 | And I had friends a few years ago, I think they fixed this,

02:23:06.740 | who would pay for the paper and then they'd go online

02:23:09.380 | and they'd get up charged for the digital

02:23:11.900 | and there was no break, there was no connection between them.

02:23:15.420 | But publishers are not that technical

02:23:17.300 | and they can't all get you to subscribe,

02:23:19.620 | you can't have a thousand subscriptions.

02:23:21.340 | So for a long time people talked about micropayments,

02:23:23.500 | there was Blendle and the other ones which came to the US

02:23:25.900 | but didn't grow.

02:23:27.620 | And I thought if you have just a browser

02:23:29.860 | and it's protecting you

02:23:30.700 | by blocking all this ad tech tracking junk,

02:23:32.940 | it can provide you an option that uses cryptocurrency

02:23:36.660 | to let you support your favorite sites

02:23:39.460 | and even YouTube channels.

02:23:40.620 | And that we prototyped with Bitcoin.

02:23:42.780 | And that meant the user had to be of means to contribute

02:23:45.980 | and willing to contribute.

02:23:47.520 | But it could be done on the Bitcoin blockchain

02:23:49.260 | and it could be fairly efficient

02:23:50.620 | even though Bitcoin went through a period

02:23:52.260 | when we had this prototype running in 2016 into 2017

02:23:55.780 | where Bitcoin was very congested and very slow to confirm

02:23:58.820 | and the fees got very high.

02:24:00.180 | And a lot of users who were not Bitcoin maximalists

02:24:03.580 | or even experienced, we helped them out

02:24:05.740 | by embedding a Coinbase buy widget

02:24:07.660 | and they had the income to buy but it was hard.

02:24:10.220 | It was like, do I buy $5 a month?

02:24:12.460 | But the fee is like 450.

02:24:14.300 | I better buy in larger batches, right?

02:24:16.460 | And they're like, I don't wanna own that much Bitcoin.

02:24:18.580 | So it became this painful thing.

02:24:21.220 | And the real idea that I had of private ads

02:24:23.840 | that pay the user a rev share couldn't be realized

02:24:27.860 | alone in that kind of system.

02:24:30.640 | In these cryptocurrency systems,

02:24:33.100 | especially with the blockchain we switched to Ethereum,

02:24:35.220 | you can have smart contracts.

02:24:36.980 | The Bitcoin system is not Turing complete

02:24:39.020 | so what you can do with the script is more limited.

02:24:41.140 | But you can still do sort of clever things

02:24:44.540 | even with Bitcoin script.

02:24:46.180 | What we wanted to do was sort of a three-sided ecosystem.

02:24:49.320 | We wanted users, creators or publishers and advertisers.

02:24:53.780 | And we wanted the advertisers to put money in

02:24:55.780 | just like they do today but without going through

02:24:58.060 | the Googles and the app nexuses

02:25:00.420 | and all these other ad tech companies.

02:25:02.000 | Because those companies take out a huge cut.

02:25:04.460 | The Guardian in the UK once did an experiment for a month.

02:25:06.980 | They bought out their own ad space.

02:25:08.820 | They put in a pound and they were paid 30 pence.

02:25:11.180 | 70% was coming out to the intermediary vendors

02:25:15.200 | they were using.

02:25:16.180 | - Wow.

02:25:17.020 | - And that's like the opposite of what the app store does.

02:25:19.260 | The app store takes 30% and gives the publisher 70%.

02:25:22.220 | So pretty broken.

02:25:23.580 | In the old days of the Superstation TBS,

02:25:26.180 | the media owner would get 85%.

02:25:29.900 | So these splits have become really unbalanced.

02:25:33.460 | And the middle players, the ad tech vendors

02:25:35.740 | are taking out way too much money.

02:25:37.500 | And they're doing something worse which has been noticed.

02:25:39.700 | They're letting not just the malware vendors

02:25:44.140 | but also the ad fraud side which fakes the publishers

02:25:47.260 | and clickbait merchants come in

02:25:51.020 | and steal traffic from good sites.

02:25:53.260 | Because once you have a certain audience

02:25:55.220 | identified at one site.

02:25:56.300 | Jason Kalkanis told me this about his experience

02:25:58.220 | with I guess it was in Gadget.

02:25:59.700 | Which site he was running.

02:26:01.800 | Once he started using an ad partner

02:26:04.180 | that was sharing his audience information

02:26:06.200 | across multiple sites,

02:26:07.400 | he saw his competitors stealing all his traffic.

02:26:09.540 | And then what's worse is the clickbait sites

02:26:11.540 | that just have much cheaper rates steal all that traffic.

02:26:15.500 | And that facilitates fraud, it facilitates fake news,

02:26:20.300 | all sorts of problems.

02:26:21.340 | So Grape blocks it and then we give users

02:26:24.080 | the ability to give back.

02:26:26.240 | And because we invented the basic attention token

02:26:28.460 | in Ethereum, we can do this three way split.

02:26:31.000 | And we can give users a share of the revenue.

02:26:33.660 | And if they want to take it out, they can.

02:26:35.500 | Now unfortunately for us and for all of blockchain,

02:26:38.220 | the regulators are saying,

02:26:39.780 | we're gonna have to know who you are.

02:26:41.300 | There's the Treasury Department's FinCEN agency,

02:26:46.300 | there's the Office of Foreign Asset Control's OFAC,

02:26:50.780 | there's the other regulators in the federal government

02:26:55.240 | that take a very dark look at things like money laundering

02:26:58.260 | and sending money to someone named Osama Bin Laden.

02:27:01.240 | So compliance starts to come in.

02:27:03.520 | And even now they're threatening for pure Bitcoin

02:27:06.080 | sending to some address.

02:27:07.800 | If you're a Coinbase, you're gonna have to know

02:27:10.280 | who's at that address.

02:27:11.120 | - Like the actual identities of people involved.

02:27:13.960 | - Now with Coinbase members, you sign up and they know you

02:27:15.880 | and they comply with the regulations.

02:27:16.960 | They're a regulated money services business.

02:27:19.460 | But if somebody's using their own self-custodial wallet

02:27:27.240 | where they have the hardware private key

02:27:29.340 | and they're not named and they want to send to that address,

02:27:32.480 | our friends in the federal government are talking about

02:27:35.080 | requiring at some threshold knowing who that is.

02:27:38.300 | - Some threshold that's unreasonable?

02:27:40.500 | - It's not that big.

02:27:41.640 | Yeah, I don't know how this will play out.

02:27:43.540 | I think crypto is here to stay.

02:27:44.820 | I think the beauty of being able to send peer to peer

02:27:47.260 | without any bank in the middle,

02:27:48.500 | without any huge wire charge and two day delay

02:27:51.660 | and all that nonsense, it's beautiful.

02:27:53.020 | And I've used it and I love it.

02:27:54.360 | But we're pragmatistic brave about crypto

02:27:56.420 | and we realize that anything like a revenue split,

02:27:59.600 | we can't facilitate without being licensed in a certain way

02:28:02.220 | and it requires knowing who the user is.

02:28:03.900 | So our default mode doesn't know who the user is.

02:28:06.460 | It instead imputes to the user's browser

02:28:09.540 | some of the revenue and allows that browser

02:28:11.820 | to steer it back to the creators.

02:28:14.940 | And we do have to identify the creators.

02:28:16.700 | But as things improve, and who knows how it'll play out,

02:28:20.460 | there should be a day when this full vision can be done

02:28:23.380 | more fully on a blockchain.

02:28:25.720 | But regulations and the practicalities

02:28:28.000 | of today's blockchains, which are not that fast

02:28:30.400 | and not anonymous over time,

02:28:32.700 | you fingerprint yourself over time,

02:28:34.780 | we do some of this with the browser.

02:28:36.260 | So one of the ideas of the basic attention token

02:28:39.780 | is to make a hybrid system that's stronger

02:28:42.500 | than blockchain alone.

02:28:43.420 | It's the browser and the blockchain.

02:28:45.940 | And the browser is this trusted endpoint software.

02:28:48.400 | It's this universal app.

02:28:49.940 | Everyone uses browsers.

02:28:51.100 | The bigger the screen, the more you're in the browser

02:28:52.980 | unless you install fat clients for things.

02:28:56.120 | I use Slack on Mac OS and it's like a browser.

02:29:00.160 | It's based on an electron framework we used to use.

02:29:02.160 | And it's just, it's not that great.

02:29:04.720 | Some of the people at Brave use Slack in Brave as a--

02:29:07.800 | - In the browser, yeah. - In the browser, yeah.

02:29:09.080 | - I use that often, yeah.

02:29:10.340 | - And I noticed on the iPad, I use apps less.

02:29:13.160 | The smaller the screen, the browser got handicapped

02:29:16.440 | by Apple and Android both.

02:29:17.960 | And it also can be slower or not have the right affordances

02:29:22.720 | to interface with the security limited APIs.

02:29:26.780 | But in principle, with the right permissioning,

02:29:28.740 | you can make the web browser just as good as any app.

02:29:31.680 | You make it be a super app.

02:29:33.380 | And that's part of our mission at Brave.

02:29:35.120 | So we want to have the economics that got captured

02:29:37.820 | by these big tech companies through tracking

02:29:40.060 | and through social networks.

02:29:41.380 | We wanna block that for your own safety

02:29:43.340 | and then let you opt into a cleaner world

02:29:45.520 | where you keep your data defended in your browser

02:29:48.020 | and you can actually realize value from it.

02:29:49.860 | So the way our ad system works,

02:29:51.340 | I mentioned it being private,

02:29:52.300 | but how does that work?

02:29:53.380 | We don't see your data at all.

02:29:55.020 | All browsers are sort of the mother of all data feeds.

02:29:57.780 | Your history, all your searches at all engines.

02:30:00.740 | Each engine sees the queries you send to it,

02:30:02.580 | but it doesn't see the others.

02:30:03.540 | But the browser sees them all.

02:30:05.260 | Machine learning in the browser that you can opt into

02:30:07.580 | can study all that in a very complete way

02:30:10.180 | and do a better job than Google does.

02:30:12.020 | Google has cookie and scripts across the web

02:30:15.220 | from acquiring DoubleClick.

02:30:16.540 | They have YouTube, they have Android,

02:30:19.100 | they have Search, which is still their big revenue lane.

02:30:21.620 | But they don't see everything.

02:30:22.660 | The browser sees everything.

02:30:23.820 | And if it can do a good job locally,

02:30:25.220 | and this is not advanced machine learning,

02:30:26.700 | this is not TensorFlow, this is like SVMs now,

02:30:29.580 | and Naive Bayes,

02:30:30.540 | then you can match intense signals,

02:30:35.060 | intense signals from those data feeds,

02:30:37.180 | the searches, the queries, the history,

02:30:39.700 | how much you're scrolling down a page,

02:30:41.700 | how much you redid a search.

02:30:43.820 | It's all blind browser algorithm.

02:30:45.780 | We don't see that data.

02:30:47.060 | And then pick the best ad from a fixed catalog per day.

02:30:50.820 | And the catalog is fixed across a large population per day.

02:30:53.420 | And it only updates once a day

02:30:55.100 | 'cause new offers come in and old ones expire,

02:30:57.620 | sometimes every week or every month.

02:30:59.620 | And that catalog, and there can be many such catalogs,

02:31:02.740 | is sold by our direct sales team.

02:31:06.220 | And so we're making an anonymous audience

02:31:08.940 | available to advertisers

02:31:10.300 | without the advertisers tracking them.

02:31:12.020 | Instead, each browser is a little machine learning system

02:31:15.300 | that's picking the best catalog entry.

02:31:17.180 | Now, the catalog is not the ads.

02:31:18.580 | Those are big, right?

02:31:19.420 | It's a video or a webpage.

02:31:20.620 | It's just the link to an edge cache.

02:31:22.500 | And there are many such edge caches.

02:31:23.800 | We're not trying to protect them from seeing your IP address.

02:31:26.540 | It's not really feasible.

02:31:27.780 | We could use Tor, but we don't yet.

02:31:30.260 | And then some keywords about the ad.

02:31:31.980 | So it's basically like metadata and a link.

02:31:34.900 | And that's what the catalog consists of,

02:31:36.300 | and that's what the machine learning picks.

02:31:37.460 | - And the machine learning is learning

02:31:39.020 | about you specifically, locally,

02:31:41.820 | in order to choose from the catalog of different ads.

02:31:44.180 | Couldn't this possibly be like a multi-billion dollar,

02:31:47.780 | isn't this taking on the Google ad?

02:31:51.060 | - Could be.

02:31:51.900 | - So like what, I mean, one question to ask,

02:31:54.620 | there seems to be some really profound ideas here

02:31:57.380 | that are different than what the internet

02:31:59.900 | has grown up to be.

02:32:01.900 | If Brave or something like Brave,

02:32:05.340 | the ideas, the fundamental philosophical ideas

02:32:07.580 | underlying Brave went out and runs 95% of the internet,

02:32:12.580 | how does that change the,

02:32:16.740 | what are the major things that changes about the internet?

02:32:19.620 | So social networks and then the creatives,

02:32:21.620 | like YouTube creators and all that kind of stuff.

02:32:23.660 | - So let's talk about that.

02:32:24.580 | First of all, if Brave gets 95%,

02:32:26.260 | I'm gonna demand a recount 'cause I won't believe it.

02:32:29.100 | I don't know.

02:32:30.060 | I think we're trying to put things into web standards

02:32:32.940 | that can be standardized across browsers.

02:32:34.980 | So the main value of Brave will be

02:32:37.380 | the trust users have in us

02:32:38.940 | and our ability to give the best deal to users.

02:32:41.300 | So 70% of the gross ad revenue we give to the user,

02:32:45.140 | and if they go through that KYC process I mentioned,

02:32:47.940 | they can take it out.

02:32:48.940 | They can also give it back.

02:32:49.940 | They can take some out, give the rest back.

02:32:51.220 | They can add basic attention tokens to give back.

02:32:54.660 | Some of them turn off the ads 'cause they just don't like ads

02:32:57.220 | but they put in $20 a month.

02:32:58.980 | I believe Zucco of Zcashfame does that.

02:33:01.100 | And that's very generous

02:33:02.020 | because the browser is just anonymously,

02:33:04.060 | based on his browsing,

02:33:05.300 | sort of keeping score on how much time he spent

02:33:07.460 | on this video, on that website.

02:33:09.500 | And if those sites verify in sort of a,

02:33:13.220 | like getting a domain certificate fashion,

02:33:15.220 | they can get paid.

02:33:16.220 | They can get part of his $20 a month.

02:33:19.220 | So that vision could go big.

02:33:21.140 | And if it does, I hope it's across multiple browsers.

02:33:23.380 | I don't know that they'll all compete well

02:33:25.940 | on the quality of the ads,

02:33:27.540 | the quality of the ad blocking and tracking protection.

02:33:30.300 | Those are subject to competition.

02:33:31.820 | It'll take a while to standardize them.

02:33:33.340 | But I think that would be a better world.

02:33:35.340 | It would have less counterparty risk,

02:33:38.180 | fewer fee takers in the middle, really just the browser.

02:33:40.700 | We're taking 30%.

02:33:42.740 | It's sort of the app store split.

02:33:44.540 | And if we get bigger, maybe we can take even less.

02:33:47.380 | Social networks, creators.

02:33:49.500 | If you look at YouTubers,

02:33:50.660 | a lot of them are the indies that are getting some size,

02:33:53.940 | are getting sponsorship deals.

02:33:56.700 | They're using Patreon.

02:33:58.860 | They're encouraging people to subscribe

02:34:01.420 | and give them regular money through Patreon.

02:34:04.300 | But that's centralized through Patreon.

02:34:06.100 | So there's censorship hazards.

02:34:07.660 | There's a 5% fee.

02:34:09.820 | What if that were a web standard?

02:34:10.940 | What if Brave pioneered it first

02:34:12.420 | and we took 3%?

02:34:14.500 | And we did it in a way that was through your browser

02:34:17.500 | so we couldn't censor it.

02:34:19.220 | - That's brilliant.

02:34:20.260 | Do you think it could be standardized across browsers?

02:34:22.820 | Can Internet Explorer come in again?

02:34:25.140 | - Yeah, protocols are easy to copy

02:34:27.940 | and they're meant to be interoperable.

02:34:29.540 | So there's a risk there.

02:34:31.100 | And the loyal users might be tricked into leaving you

02:34:34.020 | or they might, because of that distribution power,

02:34:36.180 | you might end up getting stomped.

02:34:38.300 | I don't know.

02:34:39.140 | I can't predict the future.

02:34:39.980 | I think antitrust is back on the case finally in the US

02:34:42.340 | and certainly in Europe, DG Comp is doing its thing.

02:34:45.460 | So I'm hopeful that we'll have a period of innovation.

02:34:48.860 | People were talking, like Elizabeth Warren was talking

02:34:50.700 | about breaking up the tech companies very clearly.

02:34:53.740 | Now she didn't win and I suspect that won't happen.

02:34:56.500 | But I also suspect that Google might be smart enough

02:34:58.460 | to see they should do something more

02:35:01.100 | than just put privacy perfume on Chrome.

02:35:03.420 | They should maybe get rid of DoubleClick or something,

02:35:05.340 | divest something.

02:35:06.740 | I don't know, it might happen.

02:35:07.820 | - So Brave might inspire Google to completely change

02:35:10.460 | the way they're doing things in the future.

02:35:11.300 | - They're already doing something you may have read about

02:35:13.900 | called the privacy sandbox or Flock,

02:35:17.660 | which they have this bird metaphor going,

02:35:20.020 | turtle dove, fledge.

02:35:22.780 | But these systems have been very Googly,

02:35:25.860 | kind of over-engineered and yet,

02:35:27.820 | depending on differential privacy,

02:35:29.260 | which has weakness over time, if you know how that works,

02:35:31.420 | it's kind of injecting noise to hide you in a crowd.

02:35:34.260 | But over time, an adversary can pull you out of the crowd.

02:35:37.660 | This doesn't look like it's gonna become a standard.

02:35:39.380 | Like Apple, Brave, Mozilla, we're not gonna just say,

02:35:42.700 | oh, Google, you saved us.

02:35:44.700 | You've invented the privacy sandbox,

02:35:46.060 | so we'll all just adopt it.

02:35:47.620 | Not gonna be that easy.

02:35:48.580 | It's gonna be more like pieces of what we do in Brave,

02:35:51.540 | this anonymous ad matching

02:35:52.780 | or the blind signature cryptography

02:35:55.020 | we use to confirm the ad impressions.

02:35:56.980 | That's David Choum's invention.

02:35:59.220 | That could get standardized.

02:36:00.340 | In fact, some of that is being standardized.

02:36:02.220 | Even Google's in favor of so-called trust tokens,

02:36:04.420 | which are Choumian blind signature certs.

02:36:07.460 | But they're not using them for ad confirmations

02:36:09.300 | 'cause they don't wanna blow up their own business.

02:36:11.460 | And they need to let some of the publishers they serve

02:36:14.740 | have other ad tech scripts on the page.

02:36:17.340 | And so they're kind of caught.

02:36:18.620 | And this is something I realized doing Brave.

02:36:19.860 | I thought, what's Google's innovators dilemma,

02:36:22.900 | apart from just being mature and having trouble innovating?

02:36:26.300 | It's that they have come to depend on this ad tech system

02:36:30.620 | that has all these vendors that publishers rely on,

02:36:34.980 | because publishers aren't technical enough.

02:36:36.380 | And I feel for the publishers,

02:36:37.980 | but I realized the users have to come first.

02:36:39.900 | And if you give the users a better browser that's faster,

02:36:43.700 | then you'll get enough users to give back

02:36:46.220 | or support publishers.

02:36:47.700 | The speed and the battery savings

02:36:49.420 | and the data plan savings are significant.

02:36:51.100 | There's so much bad JavaScript involved in ad tech

02:36:53.300 | that if you block it,

02:36:54.580 | you sort of chop off what's called the programmatic waterfall

02:36:57.500 | which chains a bunch of requests.

02:36:59.100 | - Yeah, that's one of the incredible things about Brave.

02:37:01.180 | I guess you're saying you should attribute it

02:37:03.180 | to the fact that the messy JavaScript, no offense.

02:37:06.580 | - No, it's good.

02:37:07.420 | It's not my excuse.

02:37:08.260 | (laughing)

02:37:10.100 | Is, I mean, Brave just feels faster,

02:37:13.900 | even then, I mean, Chrome was fast.

02:37:16.420 | One of the things that it was like impressive

02:37:18.420 | is it showed that browsers can be really fast.

02:37:21.620 | And Brave is even faster than that, which is incredible.

02:37:23.540 | - We've blocked so much.

02:37:24.380 | And it saves the network, which means data plan,

02:37:26.940 | it saves battery because the radio consumes your battery

02:37:29.420 | when it's running more to do those requests.

02:37:31.140 | And it's just stunning how many there are.

02:37:32.940 | Like some of my Google friends were like,

02:37:34.700 | oh, that's just that bad site, they'll fix it.

02:37:36.660 | And you actually do a survey of web pages

02:37:38.540 | that they're like mostly like that.

02:37:40.300 | I know Google engineers could make everything

02:37:42.420 | super efficient, but they can't,

02:37:44.260 | especially in antitrust court, do it.

02:37:45.740 | They cannot take over all the publishers and do that.

02:37:48.420 | They're trying with accelerated mobile profile, AMP.

02:37:51.740 | They're trying to pull publishers.

02:37:53.420 | They're like, oh, you poor publishers

02:37:54.580 | don't know how to make your pages fast.

02:37:56.860 | Put them on our AMP system,

02:37:58.140 | we'll give you extra placement in the search carousel.

02:38:00.620 | That's an antitrust problem for one.

02:38:02.380 | But it's also, publishers we talk to hate it

02:38:04.300 | because it degrades their brand.

02:38:05.980 | Now they look like a gig writer wrote a piece

02:38:07.980 | that's got Google's framing and AMP URL on top of it.

02:38:12.220 | And they're trying to fix that too.

02:38:13.300 | But it just looks like Google's

02:38:15.660 | borgifying all these publishers

02:38:17.060 | and they don't wanna be plugged into the BorgCube.

02:38:18.980 | They wanna build up their own brand and have loyal readers.

02:38:22.000 | So I'm in favor of giving the user's power

02:38:25.540 | to help all the publishers

02:38:27.700 | in this little platoons and the creators.

02:38:29.500 | And so we talked about Patreon.

02:38:30.820 | What about social networks?

02:38:31.940 | Well, they're inherently like search, a global algorithm.

02:38:35.380 | You're trying to find friends of friends.

02:38:37.060 | You're doing the transitive closure of a graph

02:38:39.300 | induced by this friend of relation.

02:38:41.660 | But you should own your friend relation.

02:38:45.280 | You should own your posts.

02:38:46.940 | They shouldn't be owned by somebody else

02:38:48.620 | who can take them down or censor them.

02:38:51.020 | And your friend relations,

02:38:52.000 | you should be able to find those friends on other networks.

02:38:54.180 | And so I've tweeted about this, I haven't built it yet.

02:38:56.660 | What if the browser could keep track of those for you?

02:38:59.700 | What if the browser could maybe combine Facebook and Twitter

02:39:03.460 | and you could find your friends on both

02:39:05.260 | and you could have a sort of multi--

02:39:06.500 | - So that relationship is not owned by Facebook or Twitter.

02:39:09.900 | It's owned by you through the browser.

02:39:12.020 | - They'll have terms of use and they'll say they own it.

02:39:14.020 | But if they zap you on one and you're still on the other,

02:39:16.980 | your friends find you

02:39:17.900 | and the browser could preserve a combined view.

02:39:20.820 | You could resurrect almost across networks.

02:39:23.780 | - It's something I wanna maybe quickly ask you about.

02:39:26.580 | On that front, there's been quite a lot of centralized,

02:39:31.340 | we talked about Wall Street Bets and then Robinhood.

02:39:35.240 | There's been centralized banning of different accounts

02:39:39.340 | and removing like Parler, for example, from AWS

02:39:42.660 | and this kind of overreach of centralized control.

02:39:46.780 | Is your hope that it's possible to,

02:39:49.380 | like what are your thoughts about that in general?

02:39:50.780 | Is it, and is it possible to create tools

02:39:54.500 | that give individual people the power

02:39:56.740 | to fight back against overreach of such control?

02:40:00.100 | - So we're talking about oligarchy, I do think.

02:40:02.700 | And that, if it controls a nation state, that's formidable.

02:40:05.320 | It's the tax and the police power, the military power.

02:40:08.280 | It means that you may have the Great Firewall of China.

02:40:10.880 | You may have people in China who are jailed

02:40:13.400 | because of their tweets, right?

02:40:14.920 | This is a serious threat.

02:40:15.760 | I can't minimize it or say that we'll win.

02:40:18.280 | I don't know how it's gonna go.

02:40:20.000 | But I do think, like I said earlier

02:40:21.680 | about the cunning of reason, people find ways around things.

02:40:24.040 | The internet routes around censorship.

02:40:25.880 | And this is not to endorse any particular bad faction.

02:40:28.860 | One of the things that happens

02:40:29.920 | when you try to wave the free speech flag too much,

02:40:33.040 | you say I'm not gonna censor anything,

02:40:34.640 | and you get colonized by terrible, terrible people.

02:40:37.920 | I don't care if you call them neo-Nazis,

02:40:39.840 | some of them could be doing illegal things.

02:40:41.840 | And you don't want them colonizing

02:40:44.000 | 'cause it'll ruin your reputation and destroy your business.

02:40:46.380 | So what you really want is that kind of

02:40:48.720 | user-first subsidiarity, that subjectivity.

02:40:51.560 | I want my social networks to be composited

02:40:55.320 | in some multi-social user interface

02:40:57.080 | where I don't lose track of people across networks.

02:40:59.480 | And if they leave one or they get banned from one,

02:41:01.640 | I can find them on another,

02:41:02.520 | I can still sort of thread them together.

02:41:04.080 | - Yeah, that's brilliant.

02:41:05.040 | - And this didn't happen because browsers

02:41:07.760 | got captured by the central powers.

02:41:09.880 | And why did they get captured?

02:41:10.880 | Mostly 'cause of search, and search is a central algorithm.

02:41:13.740 | So Larry Page said this too many years ago.

02:41:15.640 | He said, with search, you're giving up a little privacy

02:41:17.960 | by handing the query over to us,

02:41:19.840 | and then we'll error correct it.

02:41:21.440 | Alan Eustace used to be a Google executive.

02:41:24.320 | He said, oh yeah, we used to laugh.

02:41:25.600 | They'd all be doing typos,

02:41:27.000 | and they'd be typing the wrong word.

02:41:28.040 | And we're like, no, dummy, type that query.

02:41:29.760 | And it's like, okay, Google,

02:41:31.360 | might wanna dial back that ego a little bit.

02:41:33.040 | But yes, you do see all the queries,

02:41:35.240 | and you can improve them,

02:41:36.120 | and you can find the best results.

02:41:37.320 | And that was Google's forte.

02:41:38.880 | When we did the Firefox deal in 2004,

02:41:40.560 | Google was really good.

02:41:42.380 | And over time, SEO, which is an adversarial game,

02:41:46.240 | and Google itself buying all these companies

02:41:48.600 | and crowding its own results page

02:41:50.100 | with its own tied-in stuff, the YouTube--

02:41:53.800 | - It's the slip or slope that happens

02:41:55.280 | when you have control over these kinds

02:41:58.400 | of really important mechanisms

02:41:59.680 | as a set of data. - Yeah, monopoly capitalism

02:42:01.400 | or cartel, you get this with the Robin Hoods

02:42:03.520 | and the hedge funds.

02:42:05.320 | You get sort of the money interests take over

02:42:08.000 | and kind of abuse their power and wear out their welcome.

02:42:10.140 | So how do you get around that?

02:42:11.980 | You have to have either new land to go to,

02:42:15.440 | which some people's ancestors, not mine,

02:42:18.280 | did to found the country.

02:42:19.640 | I'm mostly Irish-German.

02:42:22.840 | You have new virtual space people go to,

02:42:26.080 | and that requires an ISP or a colo center

02:42:30.360 | or Amazon to host you.

02:42:32.120 | It requires domain name registrar who will not strike you.

02:42:36.000 | And so when Parler was taken down,

02:42:38.340 | I thought that was egregious.

02:42:40.920 | Parler, it was not well-designed,

02:42:43.040 | and I tried it out 'cause I tried all these things,

02:42:45.940 | but I didn't use it.

02:42:46.780 | And I also felt they were being unfairly scored

02:42:49.420 | for not moderating, because you can find tweets

02:42:51.240 | to this day that are horrendous

02:42:53.280 | and threaten all sorts of violence.

02:42:54.640 | Whereas Twitter, why isn't Twitter being taken down?

02:42:56.680 | But so it was very selective.

02:42:57.640 | It was the insiders who have the power

02:42:59.920 | are gonna take out the newcomer.

02:43:01.640 | And it looked bad, sort of like the hedge funds

02:43:05.360 | shorting GameStop.

02:43:06.760 | Looked bad.

02:43:08.040 | You're seeing a piece in Time Magazine this week

02:43:10.040 | that's basically saying,

02:43:11.280 | "Yeah, we interfere with the election,

02:43:12.600 | "but it was great, aren't we good?"

02:43:14.600 | I don't know if you've seen this piece yet.

02:43:16.600 | If you tried to say that as a Trump supporter in November

02:43:20.760 | after the election, you'd get banned from Twitter.

02:43:22.880 | But now Time, in its Twitter account,

02:43:24.760 | is saying, "We saved the day.

02:43:26.840 | "It's AFL-CIO and big business,

02:43:30.160 | "the Better Business Bureau got together

02:43:32.040 | "and kept Trump from spreading fake news."

02:43:34.960 | So the country's kind of broken.

02:43:36.840 | I don't know how to fix that.

02:43:37.800 | The oligarchs have run wild, in my opinion.

02:43:40.160 | And big tech is in the antitrust dock.

02:43:42.760 | What's gonna happen?

02:43:43.720 | I don't think they get out.

02:43:44.760 | I think some of the DOJ, and certainly the state cases,

02:43:47.880 | 'cause they're separate cases,

02:43:49.080 | are not gonna go away just 'cause

02:43:50.360 | somebody got elected differently.

02:43:52.360 | And these are career prosecutors,

02:43:53.840 | and they have a strong case.

02:43:55.880 | And Google's smart.

02:43:56.840 | Microsoft almost got split up, right?

02:43:59.560 | The judge, Thomas Penfield Jackson, he overreached.

02:44:02.720 | He didn't hold a hearing about the remedy.

02:44:04.960 | He just said, "I'm gonna break you up."

02:44:06.760 | And Microsoft appealed, and the higher-level court said,

02:44:10.440 | "Go back and figure this out.

02:44:11.500 | "You're not breaking them up.

02:44:12.340 | "You didn't even hold a hearing."

02:44:13.640 | And when they got back, Microsoft said,

02:44:15.340 | "Let's settle, let's settle.

02:44:16.400 | "We don't wanna get broken up."

02:44:17.660 | Because Jackson was gonna make the OpsCo,

02:44:20.500 | the operating system company,

02:44:21.580 | and the AppsCo, Office, Word and Excel.

02:44:24.740 | And that would've been a huge blow to Microsoft.

02:44:27.540 | - But ultimately, I don't know if you're optimistic

02:44:31.100 | or cynical about the possibility of breaking up big tech.

02:44:34.500 | To me, I'm optimistic that tools like Brave,

02:44:39.500 | I love the idea of owning your friendships.

02:44:42.500 | Users more and more owning the stuff is the only real way.

02:44:46.420 | Unfortunately, it's like the Wall Street Bets subreddit

02:44:48.540 | is the only real way to fight the centralized power.

02:44:51.100 | You can't break 'em up through regulation.

02:44:52.700 | It's very difficult.

02:44:53.540 | - Certainly, I don't wanna wait for the law.

02:44:55.220 | Netscape was long dead or acquired by AOL

02:44:57.660 | and effectively dead.

02:44:58.480 | It was only Mozilla that returned Firefox to the market

02:45:01.980 | by the time that the US v. Microsoft case

02:45:04.140 | was finally settled and the penalties were put in place.

02:45:07.260 | And yet, antitrust has a role to play.

02:45:11.020 | Those penalties caused Microsoft

02:45:12.660 | to kind of turn away from the web.

02:45:14.220 | They did Windows Vista, and they thought,

02:45:16.020 | "The web's too painful.

02:45:16.860 | "We got punished in court, and we had to standardize things

02:45:19.680 | "with those icky standards people."

02:45:21.300 | So they ran back to proprietary lock-in,

02:45:23.540 | and Windows Vista flopped.

02:45:24.740 | It was late, it was bloated, longhorn, remember?

02:45:27.800 | Now, what I was gonna say, but Google's smart enough,

02:45:30.420 | they won't get split up.

02:45:31.720 | They'll split something out to get off the hook, I think.

02:45:34.920 | - This is a complicated subject, but I, myself, was so,

02:45:39.340 | I decided to journey out from the world

02:45:43.300 | of being a researcher at MIT,

02:45:45.340 | and potentially doing a startup myself,

02:45:48.460 | and I've been thinking of,

02:45:50.580 | I wanted to come to Silicon Valley to do so.

02:45:53.700 | It's the land of the entrepreneur.

02:45:56.020 | And there's a lot of my friends,

02:45:58.020 | a lot of them are successfully,

02:45:59.660 | have been entrepreneurs themselves,

02:46:01.420 | have said, "Do not come to Silicon Valley."

02:46:03.700 | You've started, you ran amazing teams of engineers.

02:46:09.180 | You started a lot of successful businesses.

02:46:12.180 | I wondered if you could comment on

02:46:14.940 | why a lot of people are leaving California.

02:46:17.220 | Is there something that could be fixed about California?

02:46:19.860 | If you were starting a business today,

02:46:22.780 | would you consider somewhere else,

02:46:24.840 | like Austin, or some other place?

02:46:27.380 | Or is Silicon Valley still, is it just a little lull,

02:46:30.740 | everybody's being overdramatic

02:46:32.380 | during this particular year of the coronavirus, and so on?

02:46:35.900 | - I think, you know, even Austin's getting overheated,

02:46:38.260 | I hear, and I've had relatives and friends move to Texas

02:46:42.460 | within the last few months.

02:46:43.300 | So Texas as a whole is a big place.

02:46:45.860 | And, you know, people are moving to Florida,

02:46:48.480 | there's a big movement toward Miami,

02:46:50.700 | Peter Thiel, Keith, all these people.

02:46:53.260 | The mayor has been very business-friendly about it,

02:46:56.260 | which I think is just good politics.

02:46:58.620 | America is fundamentally a commercial republic,

02:47:00.660 | so you would think this would be what's happening.

02:47:02.540 | For a long time, California was the golden state.

02:47:04.420 | I came here in late '76 when I was a teenager.

02:47:06.820 | It's in crushing debt due to the lockdowns.

02:47:11.380 | It's got the highest taxes, that's gotta matter,

02:47:14.500 | people will do high taxes.

02:47:16.460 | It's got, likely fires every year because of the deadfall.

02:47:21.580 | It's not global warming,

02:47:22.780 | it's because the forests weren't managed

02:47:24.220 | like they had been in the first part of the 20th century.

02:47:26.860 | Just, I would say corruption at all levels,

02:47:31.120 | especially up to the governor,

02:47:32.260 | who, you know, famously was eating at the French Laundry,

02:47:34.460 | and claimed that the outside was inside,

02:47:36.740 | and they were out in masks off, and it was great.

02:47:39.520 | Do what I say, not what I do.

02:47:42.300 | Rules for thee, but not for me.

02:47:44.740 | When you see that in leadership,

02:47:47.580 | people either run, or they get rid of the leadership,

02:47:50.400 | so there's a recall drive,

02:47:51.420 | which is about to reach the threshold,

02:47:53.500 | or in the old days, they get their guns, right?

02:47:56.140 | You don't put up with this junk.

02:47:57.580 | - But ultimately, the thing that made Silicon Valley

02:48:00.700 | a special place, it gave freedom to young kids,

02:48:04.580 | entrepreneurs, young minds, brave minds,

02:48:08.140 | to think bold, to try different stuff.

02:48:10.780 | I mean, even if the taxes are high,

02:48:12.900 | so outside of financial stuff, outside of all of that--

02:48:15.500 | - Housing's super expensive.

02:48:16.940 | - Housing's super, so it's hard.

02:48:18.780 | Okay, everything about startups is hard.

02:48:19.620 | - Pencils narrow, and they didn't plan the roads, right?

02:48:22.820 | - Yeah.

02:48:23.640 | - They got rid of public transportation in LA,

02:48:25.980 | like the Who Framed Roger Rabbit cartoon show.

02:48:28.220 | They used to have trolley cars in Portland, too.

02:48:30.820 | The oil companies and the DOD conspired to build highways

02:48:34.260 | and make cars dominant, and the rights of way are long gone.

02:48:37.380 | Like Elon's gonna go underground.

02:48:39.700 | And I rushed him well.

02:48:40.540 | That's probably the only way to do it now.

02:48:42.740 | - But is it still a place, do you think it's possible

02:48:45.700 | that Silicon Valley is still a place where magic happens,

02:48:48.140 | where the next Google's built, where the next,

02:48:50.180 | I mean, Brave is built, where--

02:48:53.540 | - I think all good things come to an end.

02:48:54.860 | I think the problem is Silicon Valley

02:48:56.620 | had strong network effects through Stanford,

02:48:58.780 | through the angel investor networks and the wealth effect,

02:49:02.260 | and originally, you have to give

02:49:04.340 | the federal government credit.

02:49:05.180 | Like the ARPANET was a government project.

02:49:07.260 | Let's not kid ourselves.

02:49:08.100 | This wasn't wild free market, libertarian capitalism.

02:49:11.140 | This was all Cold War stuff.

02:49:12.740 | You had, out of the academia, you had Shockley,

02:49:17.020 | and then the Trader SA, and Fairchild, and Intel.

02:49:20.100 | But now, when's the last fab that was built in the Valley?

02:49:25.000 | MicroUnity might've been the last.

02:49:26.380 | I don't know, I haven't followed,

02:49:27.420 | but we built a fab in Sunnyvale,

02:49:29.220 | at MicroUnity, starting in the early '90s.

02:49:33.260 | And now the fabs are overseas.

02:49:36.540 | And the one thing that I will say

02:49:38.780 | that the oligarchs have intentionally done,

02:49:40.620 | in both parties, is sort of labor

02:49:44.180 | and environmental protection law arbitrage

02:49:46.140 | by going where the labor's cheaper,

02:49:47.540 | and the environmental laws aren't as strict.

02:49:49.460 | And that's polluted the hell out of parts of China,

02:49:51.980 | but it's made things, you can make cheaper junk.

02:49:55.060 | And this is not a story that's over yet.

02:49:58.860 | So what is Silicon Valley for now?

02:50:00.700 | It's for the network effect, the brain trust,

02:50:03.260 | the who you know, the parties,

02:50:05.380 | the Stanford sort of network.

02:50:09.460 | That's fragile too, over time, I'm afraid.

02:50:11.820 | Stanford, a lot of good professors I like,

02:50:16.060 | they still filter mainly based on socioeconomic status,

02:50:19.860 | but it's kind of a skate school.

02:50:22.140 | I had a friend hired out of Harvard 20 years ago

02:50:24.980 | at Netscape, and we talked about Harvard,

02:50:28.540 | and he said, "Yeah, there's still professors

02:50:29.780 | "who are great on the curve."

02:50:30.700 | And I said, "Oh yeah, I don't think

02:50:31.620 | "they're any doing that at Stanford anymore."

02:50:33.700 | And he said, "Yo, it was shocking.

02:50:35.040 | "Some of the students got Cs and Ds, and they were crying."

02:50:37.700 | It's like, yes, that's right,

02:50:38.540 | the precious deers can't take that at Stanford,

02:50:40.820 | so they get As and Bs.

02:50:42.460 | Now, you look at China, and people say

02:50:45.060 | what you know about China.

02:50:46.260 | They, true of Russia too, a lot of math science training,

02:50:49.700 | a lot of engineering, a lot of people

02:50:51.300 | who are doing their coursework to get the As and Bs.

02:50:55.220 | So I'm an American, I'm born on the 4th of July.

02:50:59.940 | - Really? - 4th of July.

02:51:01.540 | - Yeah. - Wow.

02:51:02.380 | - And America, as I say, fundamentally

02:51:03.940 | is a commercial republic.

02:51:04.860 | You can try to make it something else.

02:51:06.380 | You can say it's the new Atlantis and mystify it.

02:51:09.060 | You could talk about it in a more, I think, correct way,

02:51:11.300 | which is 13 colonies that grew,

02:51:12.840 | and then there's a lot of local,

02:51:15.220 | or in the original design anyway,

02:51:16.780 | the Federalist Papers talk about this,

02:51:18.860 | there's a lot of subsidiarity.

02:51:20.700 | But that's been eroded over time,

02:51:23.960 | and like I say, a lot of the offshoring has hurt.

02:51:26.660 | So what happened with coronavirus?

02:51:28.860 | People working from home.

02:51:29.980 | At first it was funny, 'cause I have friends at Google

02:51:32.100 | who used to grumble that not only did they have

02:51:34.420 | to come into the office, if they joined a different team

02:51:36.780 | that was centered at a different office, they had to move.

02:51:39.100 | Or if the VA team was reconstituted in Munich,

02:51:43.080 | which it was after Lars Bach just got tired of JavaScript,

02:51:46.700 | that they hired in Munich, or they hired PhDs

02:51:48.740 | from Germany and moved them to Munich.

02:51:50.780 | With coronavirus, everyone's working from home,

02:51:52.300 | and it's like, what a relief,

02:51:53.220 | I can work for Google from home.

02:51:54.820 | But then the next shoe dropped,

02:51:56.420 | and people started asking Mark Zuckerberg,

02:51:57.820 | hey, can I move to my hometown in the Midwest?

02:52:01.540 | And he said, okay.

02:52:02.380 | And they said, oh, can I keep getting my Silicon Valley pay?

02:52:05.060 | No. (laughs)

02:52:06.620 | We're gonna figure out what your cost of living there is,

02:52:08.460 | and we're gonna adjust your pay accordingly.

02:52:10.260 | - And these colonies and these little mini experiments

02:52:12.980 | that all combine to the big giant experiment,

02:52:15.100 | I have a, I don't know, I have this vision of America,

02:52:20.020 | which the country, so I was born in Russia,

02:52:21.340 | like I said, here, and this is truly a wonderful country.

02:52:24.980 | I wasn't born on the Fourth of July, but I might as well be.

02:52:27.500 | People still flee.

02:52:28.340 | - I still, and I'm red-blooded American at this point,

02:52:31.740 | and I have a sense that we've figured it out somehow.

02:52:35.420 | If Silicon Valley burns,

02:52:36.860 | another place will come up in its place

02:52:38.820 | that even more innovation, and people will move.

02:52:42.020 | And the remote work might change fundamentally

02:52:45.460 | how we work, or it might not.

02:52:47.220 | It might just give you the freedom to then create

02:52:49.540 | many other small Silicon Valleys throughout the place,

02:52:52.220 | like Austin included, but other places as well.

02:52:54.940 | And we've somehow figured it out.

02:52:57.660 | - I think that's true, that there will be more mobility

02:53:01.660 | and maybe new places that come up.

02:53:03.420 | I don't know if Silicon Valley has passed some sell-by date

02:53:07.940 | because it did hurt.

02:53:10.580 | The coronavirus hurt, the lockdowns hurt,

02:53:12.580 | in the sense that part of what keeps things going is social.

02:53:15.860 | And so a lot of young people,

02:53:17.140 | even before coronavirus, moved to San Francisco.

02:53:19.660 | It was very strange to watch,

02:53:20.900 | 'cause in the '80s, we all lived in the valley,

02:53:22.860 | and it was less populated, and San Francisco was grungier.

02:53:26.060 | It was more like Dirty Harry in the '70s.

02:53:28.620 | But by the '90s, and Jamie runs a nightclub there,

02:53:31.660 | and he's talked about this,

02:53:32.500 | you had sort of wealthy tech people moving in,

02:53:35.180 | South of Market, fancy townhouses being built.

02:53:37.780 | And that's continued in such a point

02:53:40.060 | that it's almost like, what's the movie

02:53:41.580 | by the South African director, Nels Jodie Foster,

02:53:43.900 | up in the space colony?

02:53:46.020 | Matt Damon is the guy on the Earth who has to go up.

02:53:48.740 | Anyway, it's about the stratification.

02:53:50.500 | It's about the great inequality.

02:53:51.660 | The people in the space station have amazing medical

02:53:54.860 | auto-docs that can extend their life

02:53:56.420 | or save them, cure cancer.

02:53:57.900 | People on Earth are all suffering, ground down in poverty.

02:54:01.280 | And that sort of happened while I was here.

02:54:06.220 | You saw a lot of money drive prices up

02:54:10.660 | along the narrow peninsula,

02:54:11.700 | and the single people wanted nightlife,

02:54:13.180 | so they were in the city,

02:54:14.500 | and the condos in the city got super expensive.

02:54:16.700 | And I know even Google friends

02:54:18.660 | who are socially responsible say,

02:54:20.780 | we should have more housing built.

02:54:22.780 | We should have, yes, in my backyard, not in my backyard.

02:54:25.860 | But that's not happening, as far as I can tell.

02:54:28.420 | And from the government to the incumbent landowners

02:54:32.500 | and renters, it's just not happening.

02:54:34.460 | And that has to drive people away.

02:54:37.460 | I appreciate that people come here,

02:54:39.660 | and you should wait for the prices to monitor.

02:54:41.900 | They will, but a lot of people are gonna go

02:54:44.780 | where the prices are lower.

02:54:46.500 | - You, and sorry for silly questions here,

02:54:50.340 | but just looking back, you have created things,

02:54:54.420 | have been part of creating things

02:54:56.540 | that have transformed this world,

02:54:58.580 | the world of technology,

02:55:00.740 | perhaps more than almost anything else.

02:55:02.700 | But you're still a human being,

02:55:07.580 | and unfortunately, this ride ends.

02:55:10.900 | Do you ever think about your own mortality?

02:55:13.780 | - Not too much.

02:55:14.620 | I mean, I'm Roman Catholic, so I am not afraid of death.

02:55:19.220 | I think a lot of people who have problems with death

02:55:24.220 | are suffering from some lack of either faith

02:55:30.140 | in their transcending death,

02:55:31.980 | or maybe they don't have children,

02:55:34.180 | or they feel like, they get later in life,

02:55:37.260 | and they feel like they've missed opportunities

02:55:39.220 | to do something that endures.

02:55:40.900 | And I sympathize with a lot, because I'm old.

02:55:43.500 | I got married fairly old, so I understand all that.

02:55:46.420 | Nothing human is alien to me, as Terrence said.

02:55:49.340 | But I don't fear it, no.

02:55:54.060 | - What do you hope your legacy is?

02:55:56.260 | - Yeah, it's gonna be JavaScript.

02:55:57.700 | I think, no, I think my legacy has more to do

02:56:01.060 | with my children and their children.

02:56:03.500 | I think it also has to do with web standards.

02:56:06.420 | It has to do with things like Brave.

02:56:08.300 | The things we did with Firefox, when we did,

02:56:10.720 | you know, I'm not gonna over-celebrate,

02:56:13.060 | but I think Brave is important,

02:56:14.100 | and we'll continue to prove this in a way

02:56:15.820 | that counts for many decades to come.

02:56:18.100 | But even Firefox, whatever its future fortune,

02:56:21.420 | showed you can restart the browser market.

02:56:23.300 | This thing you said about people opting out

02:56:26.180 | and routing around, you don't need everybody to do that.

02:56:28.580 | It's more like Taleb's stubborn minorities that do that.

02:56:31.200 | It's the lead users, Eric Van Hippel's lead users.

02:56:33.620 | You can be a few percent, you can tilt the market.

02:56:36.260 | And that can be done in spite of the incumbents,

02:56:39.380 | the moneyed interests, not being in favor

02:56:41.020 | of what you're doing.

02:56:41.860 | So I think what we did with Firefox won't be forgotten,

02:56:45.300 | and it needs to be done more, and we're doing it with Brave.

02:56:47.180 | And you could argue that other projects are doing it.

02:56:48.780 | In some ways, blockchain is doing it.

02:56:50.780 | The Robinhood take down the use of Robinhood

02:56:55.780 | by the Wall Street Bets kids, similar.

02:56:59.100 | So yeah, that kind of spirit endures,

02:57:01.300 | and I think it, in some ways, it's American, right?

02:57:04.300 | It's not hard revolutionary.

02:57:06.500 | It's not trying to burn the past and destroy everything.

02:57:08.780 | It's more like we have these certain, let's say, rights.

02:57:15.140 | We have duties, too, so there's some debate

02:57:16.900 | about which comes first in American jurisprudence

02:57:19.380 | and the founding documents.

02:57:20.340 | But as long as things are working,

02:57:23.020 | we'll be like pragmatic Americans,

02:57:25.880 | like de Tocqueville described in his writings.

02:57:29.300 | But if things get too out of whack,

02:57:30.960 | for one reason or another, too unequal,

02:57:33.380 | too oligarchic and abusive, we're gonna assert our rights.

02:57:37.060 | And even a few of us can do it.

02:57:38.860 | And even in the American Revolution,

02:57:40.440 | it was the minority who fought and put their lives,

02:57:44.100 | treasure and sacred honor at stake.

02:57:45.620 | It was a bunch of people went to Upper Canada,

02:57:48.860 | I think it was called, Ontario.

02:57:51.140 | - Yeah, that's a beautiful, I mean, that is,

02:57:53.540 | at the core, what your work stands for,

02:57:55.100 | is that a few people can have the power to transform society

02:58:00.100 | with just a few radical ideas,

02:58:01.580 | with just a little bit of code, change the world.

02:58:04.580 | - Gotta do it. - And that's empowering,

02:58:05.980 | and that is the American way.

02:58:07.200 | That's why this country is, I believe,

02:58:09.340 | the greatest country on Earth.

02:58:10.420 | That's not over-rheumatized too much,

02:58:13.000 | but I think some special things have already happened

02:58:16.940 | in this country and will continue to happen.

02:58:18.780 | And that's-- - And that spirit

02:58:19.820 | can continue no matter who comes here.

02:58:22.500 | They can adopt those folk ways and that spirit.

02:58:25.500 | - Brandon, I can't tell you how much,

02:58:27.540 | I was freaking out how much of an honor it is

02:58:29.500 | to talk to you.

02:58:30.340 | You're an incredible human being.

02:58:31.380 | This is one of my favorite conversations ever.

02:58:33.540 | Thank you so much for wasting all this time with me.

02:58:35.620 | I really appreciate it.

02:58:36.580 | - Oh, it seems like a breeze.

02:58:38.020 | My pleasure.

02:58:38.860 | - Thank you for listening to this conversation

02:58:41.340 | with Brandon Eich, and thank you to our sponsors,

02:58:44.400 | Jordan and Harberger Show,

02:58:46.000 | Sun Basket Meal Delivery Service,

02:58:48.440 | BetterHelp Online Therapy,

02:58:50.200 | and 8 Sleep Self-Cooling Mattress.

02:58:52.800 | Click the sponsor links to get a discount

02:58:54.840 | and to support this podcast.

02:58:57.000 | And now, let me leave you with some words from Jeff Atwood.

02:59:00.600 | "Any app that can be written in JavaScript

02:59:03.600 | "will eventually be written in JavaScript."

02:59:06.940 | Thank you for listening, and hope to see you next time.

02:59:09.740 | (upbeat music)

02:59:12.320 | (upbeat music)

02:59:14.900 | [BLANK_AUDIO]

Brendan Eich: JavaScript, Firefox, Mozilla, and Brave | Lex Fridman Podcast #160

Chapters