Jesse Wilson: From SourceForge to OkHttp, and Why WebAssembly Beats the AI Hype

===

[00:00:00]

Matt Klein: All right folks. Welcome to another episode of Beyond the Noise Signals, Stories, and Spicy Takes the show where we dig into the stories of the people shaping the future of app-based computing with a special focus on mobile. I'm your host, Matt Klein co-founder, and CTO of bitdrift, as well as the founder of Envoy Proxy.

Each episode we'll talk with engineers, founders, and technical leaders who've transformed the way their companies build and understand what's happening inside their systems. We'll dig into the challenges, the breakthroughs, the lessons learned, and wrap it all up with their hottest takes. So let's dive in.

Today I am thrilled to have Jesse Wilson with us, who, in his own words, is a programmer who tries to balance simplicity and efficiency. He's contributed to many open source projects, [00:01:00] including AOSP, Guice, OkHttp, and Okio. He's currently doing a solo project with WebAssembly. That is a very humble intro.

So thank you, uh, Jesse, for, for those that don't know though, um, many do, Jesse has worked on or started a lot of foundational software that many of you are using on your phones and in other systems. So, great to have you, Jesse. Thank you for being here.

Jesse Wilson: Thank you.

Matt Klein: Yeah. So how I, I love to get started is you have done so much in your career.

Would love to just almost start from the beginning, you know, to learn a bit about how you got into computing. Obviously you then worked at Google and did a bunch of stuff then you worked at, at Block and, you know, you've been really, you know, you've done so much open source work and industry work. So I, I'd just love to learn about how it all started and then we can go from there.

Jesse Wilson: Yeah. Well, I, I wanna go, I'm gonna take you real back.

Matt Klein: That's great.

Do it. Yeah.

Jesse Wilson: Um, [00:02:00] in, in like 2003, maybe 2002... open source was humongous, right? It's not humongous like it was today, but we had, um, we had SourceForge and we had the Linux kernel. And, I, I got started in open source. I was working in like a, a really tiny little company building a Java swing app.

And, um, we, we built this project called Glazed Lists. And the idea was, I wanted, I wanted the iTunes UI for my boring business software. And at the time, I mean this is, this is going really far back, but at the time, iTunes was like probably the, the best example of software that's both complex and beautiful.

And it had this great feature where you could type in the search box and you'd type like BEA and it would find the Beatles albums for you or whatever, and it [00:03:00] would filter as you type. And so we built this project... for Java Swing that had that feature, but as like a, as like a, like a, I don't know, computer science student.

We didn't build that feature like in a very precise way. We built it in like a super dynamic way where, um,

Matt Klein: and sorry, this is for like a B2B app that you were

Jesse Wilson: It's like exactly

Matt Klein: working

on.

Jesse Wilson: Yeah.

Matt Klein: Right. Okay.

Got it.

Jesse Wilson: But, but like, what we built was like, what we wanted was like filtering UI and sorting and stuff.

And it was really fun 'cause we did, we, we did something where you'd start with like a, a a source list of information, maybe it was like, maybe it was like a list of, customers. And then we would transform it to be sorted and then we would transform it to be filtered. So if you type like, I don't know, um, M-A-T-T, it like shows you all the customers with M-A-T-T in their names.

And then the, the magic, the magic was like, if, if new data showed in, it didn't [00:04:00] rebuild the list from scratch, it did it incrementally. And so that project at like a really tiny little company was fun. And it, it got my foot in the door at Nike and, uh, they were using it.

Matt Klein: Sorry,

sorry. That project you open sourced at that time?

Jesse Wilson: Yeah. Yeah.

Matt Klein: So, so, and this is I think pre GitHub. So as you said, you put it on SourceForge, I guess, or something like

that?

Jesse Wilson: Yeah, I think at the time, yeah. We, we started on java.net, which is another pre GitHub, uh, thing. It, it, it started in I think a CVS repo. Maybe not a, uh,

Matt Klein: I think

people out there

probably don't even know what that is.

Jesse Wilson: Yes, yes, exactly. It, it was bad. Um, but I open sourced that, and there's a guy at Nike using it for like footwear design software.

Matt Klein: But, sorry, why, why did you open source it? Like what was the- yeah, I mean, what, what made you do it? [00:05:00] Was it just something that people did back then and you thought it'd be cool? I mean, it's like, what, what made you post the code?

Or, or I guess like, did you have to convince other people at your small company that you should

Jesse Wilson: Yeah.

Matt Klein: Do that?

Jesse Wilson: Um, I mean, there was like three people at the company, so the convincing was easy. I, I think the part that was probably the, like the, the biggest sales pitch is that the company I was, I was at, they were just like...

They were trying to find any, any software that anybody wanted to buy. And if you were, you know, a, a hospital or a police station and you needed some software, they would say, 'oh, we have software for that.' And they would fly and then they would go build it, you know, me, the me the um, freshly out of college student would go build it.

And, open source has this great character where it's a forcing function on generality. So we didn't build like a filtering thing for, you know, medical bills or something. [00:06:00] We built a filtering thing that would, would accept any source data. And by publishing it is open source, it forces us to keep our... we, it keeps us honest.

You can't take a shortcut and say, we're gonna like, have one proprietary feature that's only gonna be relevant to this, like, I don't know, hospital database or something. And so, I think like at the time probably it was like, just like that's what the cool kids were doing and I wanted to be like them.

Matt Klein: Yeah, for sure.

Jesse Wilson: So, so open source that, and, at the time, um, this is really going to date me. At the time, the web was not a place where you did interactive applications. It was like,

Matt Klein: it was all static.

Jesse Wilson: Anything interactive would happen on desktop apps. And so, Nike had an IDE for shoe design and it was written in Swing and it uses, it used this software.

And when, um, when they reached out and said, 'Hey, do you wanna come work [00:07:00] for us?' I was like, heck yeah. And so, uh, open source opened the door for me.

Matt Klein: Yep. Sure.

Jesse Wilson: And that is a extremely, like, that is, that is ex- ex- extremely like self-reinforcing, uh, feature. And so, I went off and wrote Swing at Nike, and, and the only real lesson I learned was like, like you, you write code for your business problem.

It ends up being less reusable because of this sort of like, generality thing. And also it it like, it, it sort of... ev- every piece of code that you write in open source has a, has a risk that it could survive beyond the end of the product. But anything that you write for like some, some technical product is sort of like scoped, confined to that thing.

Like I don't think that Nike is still running this IDE for shoe design, or at least they hope they're not. But, all of the software we built there [00:08:00] is still like, live in the world. And, um, and so that I guess was the first domino in my, in my open source, uh, cascade. I dunno.

Matt Klein: Yeah, I, I mean I'm, I'm somewhat strange in the sense that I spent most of my career doing proprietary software work.

Jesse Wilson: Yeah.

Matt Klein: And I, I, you know, I contributed a little bit to open source and I didn't really get into open source until I did Envoy. But I can... I think a lot of what you said resonates with me, just in the sense that for better or worse, I think open source actually forces you to do better software. And not in all cases, but like in the common case, especially when done by professional engineers, I think you cut corners less than you probably would like.

It's if you think that it's not going to live on, which is exactly what you said.

Jesse Wilson: Yeah. Almost like, like cutting corners is probably too harsh. It's like in, in, in proprietary software, you can have your blinders on and you can [00:09:00] say like, I know that we wanna build, you know, may- maybe you have some limitation and you say like, this feature is only gonna be able to load a megabyte of data at a time, and if not, I have to statically allocate more.

And that's gonna be a wasteful, and I know that my problem domain is gonna, you know, be small. So I'll just like, I'll make it static and simple and it'll solve my problem domain. And like, that's how proprietary software works. And that's why like, building it is fast and and easy, but if you do it open source, you're like, well, I don't actually know who's gonna use this and whether they're going to, you know, whether they're gonna, whether they're gonna be, you know, if they're gonna have 50 customers or 50,000 customers in their little u- ui.

And so you end up building the more general thing and that more general work takes longer, but then it becomes more reusable.

Matt Klein: Yeah, it's a, it's a really interesting problem that I think about a lot because it's true. I [00:10:00] think it's also a double-edged sword because when you start building this very general software, it can start like performing worse for certain use cases.

Jesse Wilson: Yep.

Matt Klein: Right? Or, it can start to become harder to maintain because you're, you know, handling all of these use cases. So, I don't know.

Jesse Wilson: Yeah.

Matt Klein: I, I just think it's a really interesting problem space.

Jesse Wilson: Yeah. Um, so, so I'll, I'll keep going with my, my little bio. Um, probably the most interesting thing that happened in my early career... so I was working, I was working at Nike and I was like, I don't know, it would be like, you know, 20, 23 years old.

And my, my, my expectation of the software industry was like... like, I am a dumb new grad. I'm gonna go out to the world and all of these people are like software experts and they're gonna teach me so much. And then I go to like industry and I will pick on Nike here, but they're not, [00:11:00] um, dissimilar from anybody else.

It's just that like in these companies, the availability of like, I don't know, like expert software engineers was like way lower than I expected it to be. There were a few incredibly great programmers there.

Matt Klein: Sure. Yeah.

Jesse Wilson: But for the most part, it was people who'd like... they learned whatever they learned at the time, you know, 15 years ago and they were begrudging all the technology that we... that had come since.

And so I was a little bit disillusioned and I was like sort of expecting, it's like Nike big company, you know? Amazing. I'm gonna learn so much. And I go there and like a year in, there's just like, there's like nothing left to do, um, in terms of my own career growth.

Matt Klein: Well, I'm just saying, I mean, to, to be fair, you're an exceptional software engineer, right?

Jesse Wilson: At the time. I didn't know

that.

Matt Klein: Sure. But I'm just saying that, you know, I think that, um, I think what you're saying resonates with me. I mean, I, I actually [00:12:00] started my career at Microsoft and I remember I'll tell everyone out there, I was like, petrified before I started at Microsoft. I'm like, I'm not gonna be smart enough.

Jesse Wilson: Yeah.

Matt Klein: I'm not, like, I'm not gonna be able to do it and whatever else.

Jesse Wilson: Yeah.

Matt Klein: And, um, I think that, you know, there's a, there's a subset of people that just kind of like yearn for more, like more learning, more and more, more, more. And you kind of like suck in all of that knowledge that you can get and then you want more knowledge.

Anyway, sorry. Keep going.

Jesse Wilson: I mean, I guess the,

Matt Klein: yeah,

Jesse Wilson: the, like, the positive frame of it is like, you shouldn't have imposter syndrome because everybody is like,

Matt Klein: I agree. Yep.

Jesse Wilson: Yeah. Um, so

Matt Klein: I agree.

Jesse Wilson: So, I'm writing my code and, um, I'm using, I'm gonna go really technical here. So we're using like the Java collections library and you can have like a sorted list of, um, maybe it's like a sorted list of shoe sizes or something, you know? Or a, a sorted list of, uh, shoe dimensions, like wide and normal [00:13:00] or something.

And you can, um, you can remove a, you can remove a sorted list from a non sorted list, you know? Just say like, you can take two collections and you can, intersect them and subtract them and stuff. And the Java collections API does this completely reasonable thing where if you have a collection with a thousand elements in it and you want to, um, intersect it with 10 elements, it will iterate over whatever the collection is smaller and then do contains checks on the other one.

And so, if you have, you know... that that ends up being like a, a net faster thing, not necessarily like a algorithmic complexity faster, but, um, it's a good optimization except in the super weird case I was, I was in where one of the collections is case sensitive and the other is case insensitive.

Matt Klein: Mm-hmm.

Jesse Wilson: In which case, or, or if I should say like if they use a comparator. And so what would happen is we were [00:14:00] writing our system and if you removed the big set from the small set, everything worked. And if you removed the small set from the big set, it didn't work correctly. And I was so upset 'cause I was like, my code shouldn't have bugs and like, this behavior is asinine.

So I messaged the guy who wrote the Java collections, API, um, very, fantastic programmer, sort of like a... somebody I really looked up to. His, his name is Josh Block.

Matt Klein: And at that point probably worked at Sun, right? Still? Or no?

Jesse Wilson: He,

Matt Klein: I don't,

Jesse Wilson: he, he'd recently started at Google.

Matt Klein: Oh, okay.

Jesse Wilson: And so I just messaged him and I was like, look at this like, sort of like surprising behavior.

And I didn't really want to shame him, but I was like, this is a, this is interesting. And maybe I was just looking to like, reach out to this guy 'cause I really, I'd read his effective Java book. I really appreciated his, uh, his work and, um,

Matt Klein: that

Jesse Wilson: I didn't know it at the time.

Matt Klein: That was, that

was back when people wrote books for, for like programming stuff.

Jesse Wilson: Exactly. [00:15:00] Uh, dating myself even more, and I didn't realize at the time, but like every interaction between two engineers in tech is a job interview. And I email him and I'm like, Hey, here's this bug. And he's like, cool. Wanna come interview at Google? And um, and like, that's sort of like really where, both, both like my career started to accelerate, but also just like my, my appreciation of like what the process is of this industry where like, like, getting, getting an interview at Google, especially at 2006 just seemed like so...

fantastic of an opportunity. Um, and at the time I was not thinking of myself as like interview at Google material. I was just thinking of myself as like a, a, a programmer sort of working his way up the ranks at a shoe company. And, um, I, I said yes, and I did, and I, I joined that company. And, again, at the time, that was an incredible place [00:16:00] to be because finally I found like a ton of... curious and smart engineers who were basically like, like half of them were, you know, building the future of search and half of them were just building Google Labs projects that would never go anywhere.

And that was like a very fun, sort of like workshop company to join. I joined like the, I like the most boringest, most boringest product I, I sort of saw in Google's portfolio, which is the one that employed the most people, which is like the ads team, right? And, I work on the ads team and we had this just incredibly terrible monolith and that, you know, big complicated code base with no tests that was slow, was processing, you know, like millions of dollars of Google ad revenue every hour.

And, uh, [00:17:00] that was a incredible recipe to like, you know, do software architecture well, because we had this code base that was written... like written well and written reasonably, but had grown just so much. You know, like if, if you're Google in 2006 and, uh, suddenly everybody wants to pay you to run ads, the like quick and dirty ad system you built suddenly has a hundred engineers contributing to it.

And, um, like, like nothing, nothing in the, in the ecosystem was sort of prepared for that amount of growth. And that's where, uh, Google Deuce was born as like, uh, we need something to make this code-based testable. And so, we, we, we scratched our own itch and did a dependency injection container. The person created it is, uh, Bob Lee, uh, recently departed, rest in peace,

absolutely awesome person, and Kevin Bourrillion. And they paired up and I sort of [00:18:00] was on the sidelines helping out and we built, uh, just a really cool. DI container. Um, and like, like it gets a bad rap today because it's like my dependency injector just makes all my problems, runtime problems. But, um, at the time it was like, those are much nicer problems to have

'cause we can write tests and like, again, at the time it was like, 'testing your code? We don't have time for that.' But like, even at Google and we started to build that culture there and it was really, um, really fulfilling and fun.

Matt Klein: Yeah.

So, so tell us more. I mean, so you did Guice and then

Jesse Wilson: We did Guice.

Matt Klein: Yeah,

Jesse Wilson: we did Guice in guava. And, um, the guy who started Guice, went off to go work on AOSP, which was like, again, just like, like really weird random opportunities. Like why is, why is Google doing this like mobile phone project? We're a [00:19:00] completely, you know, they were a completely server team.

You know, web, web was the, was the foundation of the company. And, um, they'd acquired, they'd acquired like the Android company.

Matt Klein: Mm-hmm.

Jesse Wilson: And that company had a very different culture to Google. Google had like, sort of got this, um... engineering rigor was the name of the game, and like, like really strict, uh,

Matt Klein: without tests,

Jesse Wilson: design processes,

Matt Klein: no tests.

Jesse Wilson: Yeah. Yeah. Well, I mean, like, that was, that, that was, that was, um, that was something that, that, you know, got very, um, aggressive very quickly.

Matt Klein: Sure.

Jesse Wilson: Uh, even things like, like, they had like, I think they still do, they had like really, really strict guidelines around code formatting.

Matt Klein: Yes, I

know.

Jesse Wilson: And so you, you get like your code formatting rules and you have to like sort of be blessed by the code formatting, uh, high priests in order to commit code. [00:20:00]

Matt Klein: I mean that was also back when I think people today take for granted that almost all code that we write is auto formatted. It's like, I haven't had an argument about code formatting in like a decade.

Jesse Wilson: Yeah, thankfully.

Matt Klein: But, but, but like people don't realize that back then, like people would knit about, you know, like indent on function signatures and all of those things. It seems just like crazy to talk about that at this point.

Jesse Wilson: Yeah. Yeah. Um, anyway, the Android team was like just a... a bunch of cowboys who were hacking things together left and right.

And, um, Bob invited me to join that team, and I did. And, um, that was really like, totally fun because we were, we were sort of, um, again, like working in, working in a project where at the time Android seemed like much more of a philosophical exercise than as a product. There was like, there was [00:21:00] like the rumors that Apple was gonna make a smartphone 'cause they had the Moto Razor, um,

Matt Klein: yeah.

Jesse Wilson: iTunes phone, and there was Blackberry and uh, I guess there was like Windows CE or something.

Matt Klein: Windows Mobile. I actually worked on Windows Mobile, but anyway, keep going.

Jesse Wilson: Excellent. Yeah, I mean, uh, but, but like the, the sort of like the, the modern generation of smartphones had not yet been birthed and, um. Oh, and there was, there was palm.

Palm was so cool.

Matt Klein: That's right. Yes. Okay.

Jesse Wilson: And so, uh, we were building our Java thing and, again, I, I got involved in that and we were, we were sort of just like trying to build, you know, Android as a, as close as possible to Java, but never call it Java platform, so that you could build systems interchangeably between, uh, server Java and Android Java.[00:22:00]

Um,

Matt Klein: but sorry, this is also just so that people can conceptualize the timeframe, as you said, you're working on this and iPhone had not yet launched, right? So it's like people had not yet seen what the competition was about to look like. So you were working on like a pre iPhone system essentially. Is that right?

Jesse Wilson: Yeah. And at the time, even systems, like, like having wifi on a mobile device seemed quite remarkable. Phones were, you know, da- you paid for data by the, uh, by the text.

Matt Klein: Yeah.

Jesse Wilson: And, uh, and we, we didn't know that like, 3G and then LTE and all this stuff was coming.

And, I saw like early demos of Android and it was really compelling 'cause like we- Go- Google maps, like again, if you, if you wanted to navigate somewhere, you would go to mapquest.com on a [00:23:00] computer and you would print out a map.

Matt Klein: Yep.

Jesse Wilson: Uh, anyway, um, that was, that was a really fun place to be and we, we really did, change the world and

another sort of just like Jesse, career opportunity. Um, the, the team I joined, we were responsible for, we were responsible for like all the, the standard libraries, like, like a Array List and String and the XML parser and the JSON parser and the Http client. And I had this one conversation where somebody from the Chrome team, I forget his name, he was working on Http two, which really didn't exist at the time.

And he and I went for a walk and he was like, I want to get Http two in Android, you know, coming.

Matt Klein: Well, and, and I say, and at that time that would've been speedy, right? Like this is, this is way

Jesse Wilson: that was speedy.

Exactly.

Matt Klein: Pre- [00:24:00] Http two. Yeah. Okay.

Jesse Wilson: Exactly.

Matt Klein: Got

it. Yeah.

Jesse Wilson: Yep, yep. And, um. I am stubborn, and I, I like, made a promise to him.

I'm like, yeah, we'll, we'll absolutely build speedy for Android. I, I know that that's like a thing that's strategically important to the company to make networking faster. And then, I made that, I made that commitment and then I left Google. Uh, I followed my friend Bob, who, uh, who recruited me to Android.

He also recruited me to join Square. And, uh, I had made this commitment that we were gonna build an Http two client for Android. And so that became, okay, Http two, where I continued the sort of like commitment of building an Http two client in Java, but I didn't do it as a member of the Android team. I did it as member of the Square engineering team.

And at the time, that was like great bread and [00:25:00] butter work for Square, where, the company basically measured its success based on how much time between a customer tapping their card or at the, you know, swiping their card at the time through a, a square card reader between doing that action and seeing a check mark on the screen.

And we were measuring that to the millisecond because we wanted to make it so that the Square experience was better than our competitors. And we also knew that if more people, if, if you have a faster credit card experience, more people will use credit cards. And Square's business is basically built on, maximizing credit cards.

Matt Klein: Yeah.

Jesse Wilson: Um, and so we built OkHttp as the Http two client for Java, but mostly Android. And again, just like, kind of like being in the right place at the right time and also having the stubbornness to like, make that thing come true, even though it was, um, [00:26:00] technically difficult. And I don't think we really had, we didn't like really have business doing it, but we did it anyway.

Matt Klein: Yeah. One thing that I would, I would love to ask about is, um, or at least one thing that I found interesting about the Android ecosystem is that there are a lot of foundational libraries that everyone uses, like OkHttp, um, but they're, but they're separate libraries.

Jesse Wilson: Yeah.

Matt Klein: They're not part of the core distribution.

Jesse Wilson: Yeah.

Matt Klein: And like, one, one thing I, I'd love to learn about, which I think you would have unique context on since you came from Google and then went to, went to Square, is... i, I, I don't wanna say like, was there conflict, but like, I would imagine over the years you had a lot of conversations with the people that were still working on Android core, right?

About

Jesse Wilson: yeah

Matt Klein: whether some of these things should be part of Android.

Jesse Wilson: Yeah.

Matt Klein: Like versus separate libraries. And I'd love to learn more about that, right? It's like, how is it that, to this day, I find it actually kind of odd [00:27:00] that like, something as foundational as OkHttp is not part of the core Android distribution and it's still a separate library.

So like, tell us, tell us more about that.

Jesse Wilson: So, I, I think a, a thing that's really important about the sort of philosophy difference between Android and iOS is Android is sort of like born in open source and it has like a open source ethos. And, Google, especially in the beginning was a, was trying to pitch Android as this is an industry effort, not a Google effort.

And so there was no, there was no like, rejection of like an open source project being infrastructure for Android. And if you read the Android sources, a ton of it is code authored by Google, but a lot of it is just aggregation of open source components. I mean, it's, it's like built on Linux and and [00:28:00] Apache Harmony and a bunch of other, um, systems.

Skia is like a really important part of Android and it's just like, you know, visible infrastructure. And then the, there's like a real technical benefit to open source in Android, which is that, the software is, is bundled with the application, not with the operating system. And that is an incredible superpower for developers and also for basically revving infrastructure faster.

And so if you're, if you're writing code on Android today, you can use the Jetpack Compose framework and you can, you can use, you know, the betas or the alphas of Jetpack Compose. And it's sort, sort of like your risk tolerance for how bleeding edge you want to be. If you're on iOS, you're stuck on whatever version of Swift UI is is available to the oldest devices that you wanna support.[00:29:00]

Matt Klein: Yep.

Jesse Wilson: And this, this is like a complete disaster for Swift UI as like a platform because, basically, it as a platform was only stable and like sort of complete and fast enough, maybe like iOS, uh, 17. And, uh, Compose has been... has, has like evolved and become as stable at the same rate as that. But everybody, no matter how old of a phone they're targeting can use the latest Compose code.

Whereas on iOS, you're still stuck using the Swift UI version that you know, your oldest users support. And again, so like for OkHttp, it's like, well, Google wants the best Http two experience, which means like the, the... they want the JAR file to ship with the apps or the, the library to ship with the apps, because then the apps have sort sort of like the maximum [00:30:00] incentive to use the latest version with the latest fixes and the latest TLS configuration and stuff like that.

Matt Klein: Well, and not, not to mention the fact that on many of the Apple components, you can't see the source code. So I mean, it's like trying to debug what's, what's going on. And I, I feel like today for most of us that work in any kind of infrastructure, whether it be phone or server related, not being able to see the code of the stuff that you're building on so that you can debug and like help, help them do better by submitting patches and all of those things, um, I have found that odd as well, you know, just from the perspective of it seems like

Apple, for certain portions of their system, could not only build a better larger community, but maybe have better products and all of those things.

Jesse Wilson: Yeah.

Matt Klein: If like more of it was a bit more open so that you could see what's actually

going on. But anyway. Yeah.

Jesse Wilson: Exactly. Yeah. Um, there's, there's nothing better [00:31:00] than like debugging through some low level thing in Android, seeing the source and then being able to say like, 'oh, this, like this changed in that recent Android version.

What do we need to do about, you know, this bug that we introduced or something.' And you can say this, you know, get blame on the operating system and say, oh, this changed in,

Matt Klein: yeah,

Jesse Wilson: you know, this year's Android release, we'll have our feature flag toggle based on that. Whereas in iOS, you sort of have to like...

bisect to figure out when did this like behavior change?

Matt Klein: Well, and not to mention the fact that if you're fortunate enough to work at a company like Square, obviously not on the app that's running on the consumer phones.

Jesse Wilson: Yeah.

Matt Klein: But like for the devices that Square/Block sells,

Jesse Wilson: yeah

Matt Klein: I mean, they obviously like own the whole thing, right?

So it's like if push comes to shove

Jesse Wilson: Yeah.

Matt Klein: They, they can patch the Android OS and they can have a Linux patch and all of those things, which makes it, as you say, faster to, to [00:32:00] iterate on problems for sure.

Jesse Wilson: Yeah, definitely.

Matt Klein: Um. So, so I guess to, you know, to, to continue on what I was asking you.

Jesse Wilson: Yeah.

Matt Klein: Was there ever discussions about, I don't know, like bringing OkHttp or something that foundational

Jesse Wilson: Yeah.

Matt Klein: Like into the Android core? Or was it just accepted that like, this is gonna be a separate library that everyone uses?

Jesse Wilson: Alright, so I'll I'll give you some, uh, some real sort of inside baseball here. So there, there was HttpURLConnection, which is like the Android built in, or I guess the Java Library sound Library Http thing.

OkHttp started life as a fork of that. And then we changed the implementation to support Http two, and then we changed the API. So it doesn't look like that at all anymore, but if you look like deep enough to see things like where our header parsing code comes from, it's from, it's from that. Google in... 2015? 2014 or something? [00:33:00] Um, imported OkHttp as a third party library in the os, but it's invisible to the public API.

And so it powers that API that it originally was a fork of and they don't really want to say we, we ship OkHttp, you know, like 5.1 in the library because then

Matt Klein: that makes sense

Jesse Wilson: you lose that unbundling advantage. But, the Chrome team is now working on Http three, uh, history rhymes, and now they're keen to get like an Http three client into Android so that all the apps can do it.

Matt Klein: Yeah.

Jesse Wilson: And we have ourselves the first like genuine, like conflict of interest or, or conflict of, of, of goals here where app developers for the most part are using OkHttp, not because it's the best API Http client today. It- I think it is, but not because of that [00:34:00] they're using it 'cause it was the best Http client 10 years ago.

Matt Klein: Sure, of course. Yeah.

Jesse Wilson: When they made that decision and now like changing their mind on it is hard. The Chrome team has a library in Android, I don't know, Android 30 or something called the Http client, and it's built on the Chrome networking stack. And nothing would the team who maintains the Chrome networking stack like better then for every app developer to switch from OkHttp to the one that's built into the platform.

But, that's a big, complicated invasive, code change. And, the, the benefits are sort of like you get the potential benefit of Http three or maybe the real benefit of Http three. But the cost is, is that the abstraction in that that built into the OS Http client

Matt Klein: Yeah. Is then

broken

Jesse Wilson: is very high level.

Matt Klein: Yeah.

Jesse Wilson: Um. Anyway, so like [00:35:00] they... I've been meeting with them and trying to find a way that we can get the Http- Http three features from Cronet

Matt Klein: Yeah.

Jesse Wilson: Available to people who are using the OkHttp API.

Matt Klein: I'm just, yeah.

Jesse Wilson: Big complicated process. Yeah, go ahead.

Matt Klein: I'm just laughing because I, I had actually joked to someone that I was gonna ask you about the issue that's been open in OkHttp for like eight years around, like, when it's gonna support Http three.

Jesse Wilson: Yeah, yeah, yeah.

Matt Klein: Um, and, and, but you actually brought it up, so,

Jesse Wilson: yeah. Um, yeah. Uh, I, I think the, the part of Http three, uh, I've, I've got like a, a, a bunch of different tensions. So, Http three is big and complicated and, the amount of code to implement is probably larger than,

Matt Klein: yeah,

it's a lot.

It's a lot.

Jesse Wilson: It's a lot.

Matt Klein: Yeah.

Jesse Wilson: And so like having OkHttp depend on a third party library [00:36:00] to, to, like, there's a library called quiche that implements, but the, the hazard of that is, is like, maybe this is the, the drawback of vendoring or, or embedding your Http client is like, now we make your application a megabyte larger for the possibility that you can negotiate an Http three session.

Matt Klein: Mm-hmm. Yep.

Jesse Wilson: And our APIs are all sort of like, they all make assumptions about how the um, TLS handshake will work. And it's different in Http three and like we painted ourselves in a corner where we sort of just like, of course it's built on TCP and, uh... that- there is, like, there's some, some hazards to getting all of that working.

I would, I would really like to make it so that, app developers who are using OkHttp's APIs today will get an Http three connection in the future without making any code changes.

Matt Klein: Right.

Jesse Wilson: But for [00:37:00] that to happen, I think we need an API that looks like TCP built on, um, QUIC or built on the, like the, the QUIC part of Http three.

Matt Klein: Yeah.

Jesse Wilson: Which is different from an API that like, looks like, uh, UDP.

Matt Klein: Right.

Jesse Wilson: So,

Matt Klein: yeah, I mean, this is actually why I was asking because I, I think this is a super, it's a super interesting topic because you're balancing a lot of different concerns. You're balancing, as you said, the ability to iterate faster on code embedded on your app, but there's considerable size concerns.

You're balancing the realities of human engineering. I think the part that you didn't say, which I know from inside baseball, is that, you know, the Cronet networking stack is like constantly defunded and there's no one working on it. And then there's like people at Google working on Envoy Mobile, and it's unclear like what they're doing and who's doing what.

And I just like, I think it's really interesting, right?

Jesse Wilson: Yeah.

Matt Klein: Because at the end of the day, um, [00:38:00] I agree with you a- as a developer on Android, I don't want to think about it. I do want that part to, to look like Apple, where I just

Jesse Wilson: Yeah

Matt Klein: open up a connection and it does the right thing.

Jesse Wilson: Yeah

Matt Klein: and I don't worry about it, right? But, but now,

Jesse Wilson: exactly

Matt Klein: we have all, we have all these dependencies that I guess it starts to leak that to me as the app developer and that, that seems bad. And that, that was why I started to ask all these questions around like,

Jesse Wilson: yeah, yeah, exactly

Matt Klein: have you thought about, you know, building it into the os and, I don't know, it's just, it's a, as I said, it's a very interesting problem.

Jesse Wilson: One, one thing that I really haven't done is like, what would, what would it look like to really just try and do it, you know, to bare knuckle it and build, uh, QUIC, QUIC underpinnings like in Kotlin. I, I have an intuition that it would be a bad result. Like I think that, um, it would not be fast. But, I don't know, like, like [00:39:00] we built our own Http two stack and it was complicated and hard, but also really fun.

And it gives us a lot of,

Matt Klein: yep.

Jesse Wilson: It gives us a lot of API power. The fact that we own our own core engine for the network calls means that we can expose observability hooks in places that, uh, Cronet does not.

Matt Klein: Yes,

absolutely. Um, I mean, at least speaking from the Envoy perspective,

Jesse Wilson: yeah.

Matt Klein: I've seen the transition obviously from Http to Http two to Http three, and I feel like it's an order of magnitude complexity increase with like each jump, right?

Http is so basic. I mean, someone who's a good programmer can write the whole thing in probably a weekend. Um, that is, that is not true like as you move to Http two and then especially Http three and even in Envoy right now, for reasons which I won't get into, they're trying to change the Http two library from one that [00:40:00] Envoy had historically used for a long time to one that's provided by Google and...

that's been a multi-year effort, right? To actually sort- and this is Google. They're running this thing at massive scale and it's like we're still finding all of these bugs. And that's on Http two, like let alone the complexity of QUIC and Http three and all of the things that happen there. So, I don't know, it's like, I do feel like there's a big tension here of... I totally agree with you,

you want that flexibility of being able to iterate quickly, but having, having a stack that you can trust that is vetted and provided by the operating system for some of this stuff. It does seem useful to me

Jesse Wilson: especially like, like Http has become sort of like the, this, like the, the baseline RPC layer of all, all networking.

Yeah, I, I think that, being, being stuck on Http two is not a stable, happy place to be.

Matt Klein: Yeah, [00:41:00] for sure. Yeah. So I guess, tell us a bit more about, you know, you were obviously at Square/Block for many, many years.

Jesse Wilson: Yeah.

Matt Klein: Um, did you spend really the majority of your time working on open source? And, I would imagine that as in many companies where you're primarily working on open source, you know, I mean there still has to be some business justification for like, what, what you're doing.

So, would love to just briefly touch on, you know, I guess like what, what were your high level goals when you were there, you know, that allowed you to do all of this foundational work.

Jesse Wilson: Yeah. I should say that a lot of- in, in 2013 we started this product called Square Cash. And for a while, I didn't do any open source because it was, it was sort of like all encompassing

you know, these, these projects that, sort of like the, the real feel of a startup [00:42:00] where, you have a very small team and you're working very, very fast and everything is very fluid and, and dynamic.

And I, I worked on the server writing proprietary backend server code for, um, five or six years. And the open source stuff that I did was almost like moonlighting.

Matt Klein: Yeah, sure.

Jesse Wilson: Not, not corporately funded work, but we, we sort of rolled it into our service and the service sort of like informed a lot of like how those things work.

I think that you get, you get better abstractions if you have distinct use cases. And so in OkHttp and Okio, we would like, we would say, you know, how are we gonna make this observability API work really well for our, for our backend services or whatever. And like, like, you know, this space, it's like if you're doing observability on mobile, [00:43:00] you have one set of ob- objectives where like you're, you're trying to track whether the client's connection itself is good.

Matt Klein: Yep.

Jesse Wilson: And then on the server it's like, aren't we DOS-ing ourselves again? Are we using our resources efficiently? And anyway, I guess I had like a lot of fun building out that open source stuff, but it was not actually like a top down go build, open source objective. And then, a few years ago, I, I did really sort of work on a pure, sort of open source platform thing.

We, uh, we built this project we internally called Treehouse, and the objective of it was, can we, can we build a system where we can dynamically deploy UI to iOS and Android without the app store process? Um, and that was [00:44:00] like a ton of fun. We um,

Matt Klein: I'm sorry, is that like your own version of React Native?

Is it like a similar idea and, and

Jesse Wilson: very similar,

Matt Klein: yeah,

Jesse Wilson: very similar idea in terms of like the problem to solve. I think like, like if you're, if you're building mobile stuff, React Native I think is just like... like aside from a lot of like platform aesthetics, I think that React Native is a sort of structurally better platform because you write your code once and like writing your code twice does not cost you twice as much,

it costs you like five times as much because you create a whole bunch of coordination overhead. And then you can deploy your code dynamically.

Matt Klein: Yes.

Jesse Wilson: And deploy your code dynamically is amazing. And again, like having worked on the server for so long, I got addicted to that and then I go work on the client.

It's like, uh, hey, the client's like [00:45:00] making a bad network call here, let's fix it. And the mobile engineers were like, oh yeah, we fixed that three months ago, but there's a long tail of users who haven't updated their apps.

Matt Klein: Yeah.

Jesse Wilson: Um, and so we wanted to fix that problem, but we didn't want to do React Native specifically 'cause we wanted to do, um,

Kotlin source language, and

Matt Klein: I see. Okay. Got it. Yeah,

Jesse Wilson: yeah,

Matt Klein: yeah. But in general, I mean, you're, you're, you're obviously preaching to the choir right now.

Jesse Wilson: Yeah.

Matt Klein: I mean, that's the entire purpose of our company is I,

Jesse Wilson: yeah

Matt Klein: I mean, it is, I think people who have primarily worked on server, which I would, you know, put, put myself there, when they start to really understand the complexities of the mobile ecosystem, mostly in terms of the long tail deployment and, and like the fact that code lives forever and the fact that you can't just fix things right away.

It is really eye-opening, I think for, for most people who have become used to the click deploy fix out there. And, you know, if you're really pushing it a couple minutes, that's, um, nope. [00:46:00] Not the, the

Jesse Wilson: thing that

Matt Klein: way it works.

Jesse Wilson: The thing I'm really obsessed with here is not the, like, the hazard of like, well, we've got this, we've got this bug that we shipped two weeks ago, and we're just like, we're just gonna have to live with it.

The thing that really... like, like hurts me in my, in my soul is like, because we can have those bugs and be stuck with them. Mobile has this really sort of like archaic QA process.

Matt Klein: Yes.

Jesse Wilson: Where it's like,

Matt Klein: it's just like the old days. Yep.

Jesse Wilson: It's like, it's like, well we, we are, we're gonna send this, this version of this code to our users and if there's any bugs in it, we need to either have feature flags that we can control or whatever.

And so we get this big complicated QI process and, um, you know just change- changing the label on a button. It's like, let's change the label on the button. It's like, well, what if it, what if it drops our, you know, conversion rate of our funnel by 5%? It's [00:47:00] like, well then we can't change the label on the button. We have to use a feature flag to change the label of the button.

Matt Klein: Yep.

Jesse Wilson: And then like every, everything in mobile costs, like 10 times as much as it, it does on the web or server.

Matt Klein: Let me ask you.

Jesse Wilson: Yeah.

Matt Klein: Um, do you, do you ever get frustrated with the app store restrictions in terms of what they allow you to do in apps? And, and again, to be clear, I understand why these restrictions exist, but in a perfect world, especially for people that are willing to do more interesting engineering, if you were allowed to like JIT code on the client, and I mean, I mean, there's so much cool stuff that could be done if that was possible.

Jesse Wilson: Yeah.

Matt Klein: And I, I don't know. I would just love, love to get your opinion on like, in general the, the mobile ecosystem and how, again, like how the app store it, it really limits some of the engineering that even is possible to fix some of the problems that we're actually talking [00:48:00] about.

Jesse Wilson: Absolutely. I mean, like, like there's like a whole bunch of business model problems where it's like, when I was working on Cash App, it was a payments product and it's like, maybe we wanna let you, maybe we wanna let you do like purely digital purchases.

But that sort of thing just doesn't work in the app store commission model. And then, we, we, we, we, we have this established pattern, which is that systems like, um, Lua, and, um, JavaScript core exist. And you can, dynamically run code as long as it's interpreted.

And that constraint is just a hundred percent fake. Like, like we have the, we have the software of technology to download and execute code safely. And Safari will download and execute code and run it at, native speed. But mobile apps are not trusted to do this, and I think [00:49:00] that it, it holds the entire industry back.

Yeah, I I am just incredibly frustrated with that. And, uh, yeah,

Matt Klein: I mean, there, there are, you know. Obviously without getting into too much of our company stuff, I mean

Jesse Wilson: mm-hmm.

Matt Klein: Obviously we, we do a lot of dynamic stuff on the client, like that's what

Jesse Wilson: Yeah.

Matt Klein: Makes what we do interesting. And just from a technical perspective, there's so many internal conversations that we have around... wow,

like if we could send binary code to the client, right? It's like, what could we do?

Jesse Wilson: Yeah.

Matt Klein: I mean, it's like we could do so much more. We could, we could at native speed, we could, you know, change how logs work and different types of matching and all of those things. And today, you know, as you obviously know, we're limited to interpreted code, which is incredibly slow, or maybe some java script ring or maybe some WebAssembly thing or something.

But it's, yeah, it's, and, and all of that has binary size implications because you have to embed the interpreter. It's like, the [00:50:00] whole thing is, is this is an area that frustrates me a lot. Like, just from a product perspective. But anyway. Yeah.

Jesse Wilson: Yeah.

Absolutely.

Matt Klein: So, did you, um, I mean we, we could talk about this all day.

Um, would, would love to learn a little more about... you obviously decided to leave block and now you're

Jesse Wilson: Yeah

Matt Klein: doing your own thing. Um,

Jesse Wilson: yeah,

Matt Klein: I, I don't know what you can share, but, you know.

Jesse Wilson: Okay.

Matt Klein: I mean, yeah, like what I, I, I guess is there anything more that you wanna share, share about your time at Block? Um,

Jesse Wilson: yeah,

Matt Klein: and,

Jesse Wilson: you know, I mean, I had, I had a ton of fun.

It was a great place to work and especially in the early days it just felt like, it felt like, like one of these, these, um, fabled environments like Bell Labs where, or this like early time at, at the Android team where just like innovation is happening and everything is permissionless and there's a lot of trust.

And so I'm looking for some new opportunity where I can have that sort of like, research lab sort of experience. [00:51:00] Right now, I'm- there's a technology I'm totally obsessed with, which is WebAssembly. And the, the thing I'm excited about for WebAssembly is that, it, it, it has that dynamic behavior that we're talking about.

But the place where, I think, the place where I think it's gonna like really sort of change computing is, on the server. And, there's a, there's a product I'm playing with that looks awesome. It's called ah Spin. And

Matt Klein: Yep.

Jesse Wilson: Instead of uploading a Docker container to a thing, you upload a WebAssembly program to a thing.

And, what I, what I think is like completely wild is that it finally, and like clearly, is going to destroy the microservices/monolith forced, forced chain- or forced choice that doesn't actually exist and is completely fake. So you'll be able to write, you'll be able to write big, complicated [00:52:00] applications with WebAssembly and, deploy them independently of how you operate them.

So I, I just am so excited about what that means for computing.

Matt Klein: Do you mean like you would have a monolith of like function components that conceptually fit together and it's up to the orchestration system to basically decide whether they get, whether they're in one process or they're split across machines or something along those lines.

So the, so the runtime of the system would basically be independent from like how you write the components?

Jesse Wilson: Yeah, yeah, yeah, yeah. Like, uh, like right now, in every single company, everywhere there's somebody saying like, we want to decompose this service into two because we want to have two different teams operate it.

Matt Klein: Yep.

Mm-hmm.

Jesse Wilson: And like, you can't share an- you can't operate a service with, with, you know, that many people. And then it's like, but then there's a performance cost of all the RPCs between the services.

Matt Klein: Yep.

Jesse Wilson: And, um, we're, we're looking down a future [00:53:00] where you can, um, I can operate the code that I own and also.

All of that code will run in the same process with like zero latency.

Matt Klein: Yeah. I, I mean, I can totally buy the vision, the, the one part of it, and maybe we have to just do a whole other episode on this because I think it's interesting is, um, I mean that, that the part of it that I feel like has prevented this vision from succeeding in the past and would love to learn from you on why it's different now, is that...

conceptually it makes sense, but in reality, there's still humans involved that write that, that write software that has really tight dependencies and that might not be abstracted in the right way, that allows for the dynamic decoupling. And I guess what I wanna ask you is how do you fix that? Like, how do you make it so that you don't, like you can have this amazing runtime, but if people, as we know they do, compile this giant [00:54:00] blob together and it's all like sharing the code and you can't actually intertwine it because there's circular dependencies.

It's like you can't fix that with a magic orchestration system, right?

Jesse Wilson: Yeah.

Matt Klein: So, it almost seems like it has to be part of a new language or like a new way of writing code or something like that. And, and like that's the part that I don't, I'm like a little hazy on.

Jesse Wilson: I think you're exactly right. Like I think it's hard and I, I don't think you can just take a program and rebuild it in WebAssembly and have capabilities.

Matt Klein: Yeah.

Jesse Wilson: Um, this is a great conversation to have with like somebody who does Envoy. 'cause Envoy is an orchestration system if you, if you squint right? Maybe, I don't know if you think of it that way, but like, um, if Http is our function call mechanism and WebAssembly can say you're making an Http call to the, the session service,

why don't I just give you the session service and you could call it in process?

Matt Klein: Right.

Jesse Wilson: So gets so, gets so cool.

Matt Klein: Yeah. No, no, no. I mean, it again, like it makes, makes [00:55:00] sense to me from a conceptual basis and I've seen different efforts

Jesse Wilson: Yeah

Matt Klein: in this area. I just, I guess I'll be honest, I'm skeptical that this is not a problem that can be solely solved

Jesse Wilson: yeah

Matt Klein: by like exciting orchestration technologies.

Jesse Wilson: Exactly.

Matt Klein: And WebAssembly, right?

Jesse Wilson: Yes.

Matt Klein: It's like there are human like, organizational issues here and how people write code and all of those things that I think have to be solved as well.

Jesse Wilson: Absolutely.

Matt Klein: But it certainly makes sense. So are you, I mean, again, like you're on your own now, so are, are you starting something new or are you just experimenting?

I mean, it's like, what is, what are you, what are you up to these days?

Jesse Wilson: Right now I am just playing with the technology.

Matt Klein: Yep.

Jesse Wilson: And, um, I am simultaneously like, extremely excited about the possibilities and a, a bit, um, humbled and disappointed by the realities.

Matt Klein: Yes.

Jesse Wilson: Where, um. Like, like the... WebAssembly has this thing called the component model, and it's just like [00:56:00] incredibly, incredibly thoughtful about capabilities and things like that, but it's also like complicated and incomplete.

Or maybe, maybe incomplete's the wrong phrase. Like it's just not done yet. There's a lot of work to do, in the ecosystem to sort of make that,

Matt Klein: yeah.

Jesse Wilson: Make that, that, uh, objective real. Anyway, I'm playing with it, getting comfortable with it, and I don't know if that will be, I don't know if that'll be what I'm working on, you know, 24 months from now or, uh, or if I will retreat

Matt Klein: Yeah.

Jesse Wilson: And just go build more mobile stuff.

Matt Klein: Yeah. I mean, I'll say that this is refreshing. You're the first person that I've had on the podcast that didn't, you know, start talking about artificial intelligence. So, I mean, there are, I'm just saying like, there are other interesting problems in computing

Jesse Wilson: Yes.

Matt Klein: That are not artificial intelligence.

Jesse Wilson: Absolutely.

Yeah.

I'm just gonna say like, I am so much more excited about WebAssembly than I am about LLMs.

Matt Klein: Well, that, that is an [00:57:00] interesting note to end on.

Jesse Wilson: Yeah.

Matt Klein: And I, I think there'll be more for us to discuss in the future, but I, I, I guess, within the computing space, you know, we are obviously talking about, you know, WebAssembly and that's what you're excited about.

Are there any other technologies and things that you think are gonna be interesting over the next few years, whether it be in mobile or other areas?

Jesse Wilson: Having done mobile for a very long time, I'm incredibly excited about the web. I feel like, I feel like the last time I really seriously interacted with the web... browser incompatibility was complicated and, everything was difficult.

And, part of this project I'm working on right now, I'm doing a lot of web stuff and it's just, it's just really cool what you can do with stuff like CSS animations and, all, all of the sort of like minimal things you want to write a really compelling app are available on the web. And then maybe one thing that's sort of tangential to that is I'm, I'm really excited about this Servo project, which will be, you know, it's, it's super [00:58:00] early, but a new web, a new web rendering engine that could power a new browser is happening and it's, um, technically super interesting.

Matt Klein: Do you, do you think then... see it, it seems like if, if you look at it from the mobile side of things

Jesse Wilson: Yeah.

Matt Klein: You know, people initially did a lot of mobile web apps, like progressive web apps, and then everyone was like, oh, we're going to native, you know, but then it has all the problems that we've been talking about, about the slow deploy speeds.

Do you, do you think that's wrong? Like, do you actually think that we should be doing more web apps on mobile? Like if that was better?

Jesse Wilson: Yeah. So, if you look at, if you look at your Mac's desktop, 20 years ago, we started this conversation I was talking about iTunes and how excited I was about iTunes. There's just, there was so many rich desktop apps that were like Mac only.

And I love like, sort of like the Mac-est Mac apps, um, idea. And if I look at my Mac today, it's like a hundred Electron apps [00:59:00] and, and a web browser. And like my music player is an Electron app. My communication thing, my email thing, it's all Electron. And I believe that like the desktop as an application platform is basically dead.

And that's fine. Like the, the web is just a great place to build, desktop apps.

Matt Klein: Yep.

Jesse Wilson: And I think that the same could be true of mobile if we, the engineering community do our work and, and like try and bring that future forward. I think that like, there's, there's almost nothing that the, the web technology can't do unless it's been forbidden from doing so.

Matt Klein: Right.

Jesse Wilson: And, even though I think there's still a lot of really exciting space to explore,

Matt Klein: yeah,

Jesse Wilson: like, like web apps are, are absolutely as fast as native apps for anything you really wanna measure. And the, the reason that we're like, obsessed with native apps, sort of like as an industry is because there was a very long time where they were dramatically better.

And I just think [01:00:00] that like the web has caught up.

Matt Klein: Interesting. Well, I think, I think that's a fantastic, place to end. So, thank you, Jesse. This was a, a great conversation. Would love to keep up with you on what you're building and maybe have you come back on and tell us, more about it. So

Jesse Wilson: yes.

Matt Klein: Anyway, thank you. That's a wrap for this episode of Beyond the Noise Signals, Stories, and Spicy Takes. Huge thanks to Jesse for joining and sharing his story. You can find this episode and all past ones on the bitdrift YouTube channel. If you had fun, drop us a review, tell your friends or yell your favorite hot take into the void and just make sure to tag us.

I'm Matt Klein and I will see you next time. Thanks again.