3 ms·
OK, let's break down and analyze the authors arguments which IMO, are good, but perhaps are more of a jumping off point than a conclusion. He argues that web te
by mike_hearn 4y ago
OK, let's break down and analyze the authors arguments which IMO, are good, but perhaps are more of a jumping off point than a conclusion. He argues that web tech tends to leak out of browsers and into other domains. There are lots of examples of this. This occurs because:
1. Web tech is hard to change after launch due to the sheer size of the web and the amount of abandoned content, so it remains stable and is designed to be general. Developers like both.
2. Browser tech is subsidized by ads and hardware sales so has big budgets yet gets given away for free.
3. Browser makers (or rather, Google/Mozilla) insist on patent-free technologies. This can potentially kill off or stymie the development of more advanced tech, in particular in the codec space, but it's convenient.
He then argues that WebAssembly, WebGPU and QUIC will be the next technologies to go the way of JSON, XML etc and get widespread adoption outside the browser. And that's about it. The author passes little judgement on this trend beyond observing that it happens.
There's definitely a lot of reasons why our industry has settled on (let's face it) abusing a document format and document renderers as a way to distribute software, and Paul identifies some of them. In particular the rigorous commitment to backwards compatibility in a way that's PM-promotion-proof is hard to beat. It comes from a combination of the enormous amount of free content that any new browser (version) wants to be able to render, along with the network reachability of that content allowing very solid regression testing programmes. Other platforms tend not to test against their user's codebases in the same way that browser makers do.
Still, there's a lot of dissatisfaction with web tech out there because it was simply never meant for what we do with it and the seams are really showing these days. The sheer number of pre-processors and languages-transpiling-to-other-languages that modern web tech involves is a testament to that.
It feels like there's scope for competitors to the web stack to arise. I've noticed a lot of developers like to spin implausible stories about the importance of being on the web and how "nobody downloads apps anymore", but there's not much concrete evidence for this. Mobile/tablet is obviously a major hole in this theory, along with the fact that people actually download Chrome itself, the success of Minecraft, the existence of the entire video games industry etc. People often end up over-betting on web tech because they like the idea of a standards based platform, but then quietly hit its limitations and give up (most notable example: Steve Jobs). I remember when WebGL was new there was a wave of predictions that video games would all be web apps in future. Yeah, didn't happen.
What would a competitor want to emulate about the web? There's a million things that could be written about what the web does well and badly, but to focus on just one issue raised in the article you'd really want a way to discover downloadable apps that were built with the platform, so you could download them and run them against new versions of your platform. Once you can precisely measure how much stuff your new version breaks you can then set metrics around it and drive it to zero, so developers get a rock solid platform with no regressions. It'd also let you quantify the cost of breaking backwards compatibility (which should discourage it).
You'd probably also want a much higher degree of modularity. Although the article claims web tech keeps leaving the browser and being reused this is really more in spite of browser makers, not because of them. The Chrome team don't actually care about anything they do being used outside of Chrome itself, which is why stuff like Electron and NodeJS all come from external organizations, and why there's tons of useful tech in the Chromium codebase that languishes unused. It's not well documented, it's often not particularly well modularized, etc. Any competitor could certainly do better and provide more like an IKEA kit of components that, in one instantiation, happens to provide a browser like experience with sandboxing, tabs, code streaming etc, but in others could be more like an Electron-style library experience.