The enduser couldn't care less what if-else nightmare logic you ran in your "application engine" in order to reach the current framebuffer, they care only about the framebuffer. Following this logic, that tidy codebase should perhaps be also tiny: in the extreme case, just thinking on a whim right now, it should be comprised of only one "module": a stable diffusion-like algorithm which ingests the HTML/CSS/JS specification documents and the requested website response and "simply" outputs the adequate framebuffer, with which the enduser interacts accordingly. This kind of approach probably reduces the cost from billions to only a few millions for the GPUs training time. Sort of a "generative browser", a "genser", if you will.
The if-else nightmare logic it's exactly what drives developers away and a surefire way to kill a project before it's even born.
Since the success of Ubuntu as a distro, people started to think of the FOSS ecosystem as a competitor to commercial software but gratis. And IMHO that's a wrong approach to Open Source.
We should go back to the roots, forget about competing, focus on what's the best not for the end users but on what's fun for the FOSS developers, and bring back the hobby on open source projects. Maybe someday, this project will become huge. Maybe not. Dunno.
I am all for going back at the roots, but at the root of all the roots is the enduser: the machine must do something useful, otherwise it's a niche postmodernist art contraption (not that there's anything wrong with that).
I don't entirely agree with the end user being the root of all, at least in FOSS. While they shouldn't be alienated, the focus should be on having a good approach to the project, one that is friendly with the idea of having to code after a full time job, like most open source developers do.
Keep in mind that this isn't a job, nobody will fire you and most complains by end users can be ignored with no consequences. If the project is not fun and engaging, then what's the incentive to keep going?
I prefer Blender's approach of heavily prioritizing users over developers [0]. It's been a huge success.
Chromium isn't successful because of the billions thrown at it. It's successful because it was significantly better than anything else during its rise:
It was significantly faster than IE and FF. Much better memory management and crash handling via separate processes. Strict adherence to web standards. Seamless auto-updates that required no user intervention - ever. No admin privileges needed for updates. Clean UI that stays out of your way. Top-tier developer tools built-in. Very secure. Has there ever been a widespread instance where users were infected with malware from a Chrome exploit? I haven't heard of one yet.
Any competitor to Chromium needs to be significantly better than it for genuine reasons. Unfortunately, "not Google" and "privacy" aren't going to cut it for most average users to switch over. Unlike Meta, Google's reputation isn't in the trash so people still trust them.
I think you're dramatically underestimating the importance that advertising and bundling had in the rise of Chrome. Every non-techie I've talked to about this basically uses Chrome because Google told them it was the best, and Google is the first internet page they go to whenever they go to the internet.
IIRC some Firefox devs also accused them of tweaking sites like Youtube in particular ways that only affected competing browsers and made performance worse comparatively.
We've seen Microsoft throw money and bundling with IE and Edge, and it still hasn't done much. Even with being able to bundle Edge as the OS default - the biggest advantage anyone could ask for.
Then Microsoft bundled IE4, and killed Netscape (it took a while, but the unstoppable momentum was built with bundling/embedding).
Chrome is “bundled” with Google. Every Google search recommends it, and everyone uses Google. Same for YouTube which to this day works better and faster on Chrome. Android (80% user base at the time it happened) also pushed Chrome.
Chrome wouldn’t have become popular if it wasn’t good. But the market dominance did not come from being good - it was just a necessary condition for market dominance in a market already dominated by incumbents.
Maybe not widespread attacks, but the Chrome team regularly see 0days being actively exploited in the wild, I imagine as rare and isolated incidents instead of mass pwning billions of users. For some exploits to work you have to visit a carefully crafted page that has the payload in it. And that's not easy to do at scale. You'd have to cajole millions (billions?) of people into navigating to a specific URL. Also I imagine you don't hear about these exploits because the actors would have good op-sec and keep all their data gathering secret.
Not even Google started from scratch. They started from WebKit and used their billions to overhaul, maintain and change everything from the engine to the renderer.
> But maybe you don't even need to build another Chromium.
That's what Servo said years ago. I don't see the progress or hype around that anymore.
Maybe Ladybird is different, but overall I'm very skeptical; but an alternative browser that is not Firefox, Chrome or Safari needs a high multi-person and project contribution for it to work.
And WebKit didn’t start from scratch either, it started from khtml.
A few years ago there used to be 4 active browser lineages, now there are just 2.
The Mozilla team didn’t intend Servo to be a fully functional new browser. The project was amazing as a Rust-only implementation of the basics of a browser. But ultimately, the utility of Servo was that it served as a place to rewrite modules into Rust for use in Firefox.
Incidentally, there were some good talks by the Firefox team about how they chose to select which modules to rewrite in Rust (since a complete rewrite with feature parity is prohibitively expensive).
This is not accurate. Some parts of Mozilla thought that way, other parts didn't. And thoughts changed over time. For years Servo was officially an experiment, with many possible futures.
Had it succeeded in rendering the modern Web well enough, I think it could have become a fully functional new browser. Sadly, that didn't happen.
(source: I was at Mozilla at the time, and adjacent to the Servo people, who I talked to a lot)
I don't want to build another Chromium. I want to build a web browser. I don't think every browser is required to be like or support as many websites as Chromium does. Note that competing with the popular browsers of today has never been stated as one of Ladybird's goals.
Mild nitpick but "web browser" and "rendering engine" and "JavaScript" are different.
Anybody can take WebKit and do some cool stuff on top in the UI and call it a "web browser".
At the end of the day, without patching to the HTML/CSS rendering engine, it's going to perform like every other WebKit based browser, right?
https://en.wikipedia.org/wiki/Comparison_of_browser_engines
I know Chromium uses Blink and not WebKit now.
Looks like Ladybird is based on https://en.wikipedia.org/wiki/SerenityOS "LibWeb"
That's all just rendering. You need to hook a JavaScript engine up to the rendering engine as well (with glue to the DOM from what I know?)
Chromium is Blink + V8, right?
Does Ladybird SerenityOS Libweb also do JavaScript?
> Browser with JavaScript, WebAssembly, and more (check the spec compliance for JS, CSS, and WASM)
It does. https://github.com/SerenityOS/serenity https://github.com/SerenityOS/serenity/tree/master/Userland/...
This is akin to saying that if you want visit the US from Europe you must built a wooden tall ship, secure a crew of sailors, pack several barrels of limes, and spend weeks sailing across the ocean.
That's how they got there the first time but the choices they made then were determined more by the information they had at the time (or lack thereof) and the technology they had (or lack thereof).
Re-implementing a browser, starting today, is a fundamentally different process from building one starting over a decade ago while the web was constantly evolving.
Of course, the n-th time is cheaper, easier, faster, case in point: I implemented 'deon', a notation format for structured data [1], using your amazing "Crafting Interpreters" for which I paid nothing since I was reading the web version as you were writing. Never had the chance to say thank you, somewhere in my drafts there is an email of appreciation: reading your book and applying it chapter by chapter, crafting a final, useful artifact, has been a beautiful experience, thank you very much, for all your writing, since I am a longtime reader of your technical and otherwise texts.
Sorry, yes. I didn't intent to disagree with your entire comment, but just remark on the first sentence of it.
I'm glad you enjoyed the book! :D
Reminds me of: "According to all known laws of aviation, there is no way that a bee should be able to fly. Its wings are too small to get its fat little body off the ground. But the bee doesn't know that, so it flies anyway."
I guess Kling & co. didn't get the memo :)
haa, that's a fun idea! setting aside efficiency, though, neural networks aren't usually Turing-complete, so arbitrary JS isn't going to work. but, I could imagine building a very strict, minimalist browser engine (think XHTML and Scheme rather than HTML and JS), and learning a transformation between the two.
and for perf to be attainable, rather than an NN you could learn a bunch of syntax transformation rules between the two.
I wouldn't worry about performance: Nvidia breaks world records with H100 [2], Intel is going for 6 GHz processors [3], for performance you just have to be patient.
[1] 2019, Jorge Pérez et. al, On the Turing Completeness of Modern Neural Network Architectures, https://arxiv.org/pdf/1901.03429.pdf
[2] https://blogs.nvidia.com/blog/2022/09/08/hopper-mlperf-infer...
[3] https://www.tomshardware.com/news/intel-teases-8-ghz-raptor-...
It would be interesting to see if a search engine's worth of raw data, and enough training on Chrome's output, could build a JS interpreter. I'm skeptical but don't see why not in principle.
The comparison doesn't make sense.
Linux was not released in 2022, competing against 2022's Windows, it was released in 1991, competing with the technology at the time, and has since evolved and grew progressively, as the competition evolved and grew.
Releasing a new browser in 2022, competing against 2022's Chrome and Firefox, is a way bigger task (not that it's their goal).
The competition at the time was also cutting edge at the time, the result of billions of dollars of investment over decades, and Linux was nothing. There's no reason that your argument couldn't have been made in 1991. Why, specifically, would it have been a bad argument in 1991, since we now know it would have been?
I don't deny that, but just the feature gap was a lot less compared to what it is now.
We're talking about a time where video games were made by teams of 5 people, not 500.
If the cutting edge technology is using a rock on a stick and you're using a rock, it's easier to catch up than if you're using a rock and they already have tree cutters working on nuclear energy.
Related: are you trying to make an argument that a small team or a single person shouldn't try to write a video game in 2022?
I don't want that at all. Firefox lost their spine long ago, and at this point they aren't a Chrome clone, but its pretty close.
My understanding is that this is exactly the opposite of how Linux was started. It started off as a barely working product on 386 only and went from there.
It was a barely working product, but it was accessible enough to other developers to develop on it and extend it.
If we start slowly, focus not on delivering a working product ASAP but a nice codebase that's accessible for other developers to read, carefully split modules so each part of the browser is a mini project by itself, with clearly defined business logic, outward connections (GUI, APIs, Drivers, etc), it can be done.
Yes, it will be a lot of work. But with a tidy codebase and a welcoming community (at least for a programmer) that "a lot of work" will instead become a playground to hack whatever component they desire.
The problem is that people don't want another Linux-like hobby project, they want another Firefox