The Web is Too Big, or Scaling Down

BY SCOTT RICHMOND

In October of 2024, Mozilla, not-for-profit publisher of the Firefox browser, announced that it's going to become an adtech company. This is real bad. Apparently somehow it's going to be a nonprofit adtech company? (Although Sam Altman seems to know how to convert nonprofits to for-profit concerns; Elon Musk has other ideas.)

Firefox has been, despite recently only being a very minor browser (estimates are not much higher than 5% of market share), an important part of the infrastructure of the web. Mozilla at least theoretically has the public good as part of its mission, in a way that Google and Apple and Microsoft do not, and cannot (under the capitalist regime as currently constituted).

And so, Firefox is a noncommercial browser made my a not-for-profit organization which has the wellbeing of the web at its heart. Mozilla, it seems, has made the calculation that the web is now fully an advertising platform, and that focusing on the public good of the web demands lubricating less-awful advertising practices. I mean, that's not exactly wrong; the web, having been subject to Google's extractive practices over many years, has been remade as a medium for advertising.

My impulse, of course, was to reach for noncommercial, non-ad-tech alternatives to Firefox (while still using Firefox). I know I'm not the only one waiting for Ladybird to get usable, and until then, there's Zen and LibreWolf and a few other alternatives. As soon as I read about Mozilla's plan, I ran brew install zen-browser. And, after about a month, went back to Firefox; Zen is Firefox-based, and doesn't actually go deep enough to fix the problems I have with Firefox (while introducing a few of its own).

Ultimately, though, the problem with this situation isn't that Mozilla or Firefox aren't good enough. It's that the web is too complicated.

Building a better browser?

A browser is an extraordinarily complex piece of software. So complex that, until Andreas Kling started working on Ladybird (originally as part of his SerenityOS project, now as his major project), it seemed rather impossible to implement a browser from scratch in the 2020s. You needed to be start in, like, 2000, writing Konqueror (which would eventually become Chrome). The web was simpler then; building new browsers has gotten less tractable as web standards have gotten more complex.

I am, of course, not the first to make this point. Drew DeVault argued as much in 2020. He was very persuasive. I can't recall where, but I have seen DeVault offer a mea culpa once Kling's work seems to have taken off. And Kling himself says he was in part inspired by DeVault's doomsaying. But Ladybird is not fit for use, not yet.

It's hard to overstate just how unavoidable the complex, "modern" web is. You might not expect it, but Google Chrome is the standard platform for writing desktop programs these days (in the form of Electron). Spotify, Discord, Microsoft Teams (🤮), Signal (💜), and Obsidian are all built with electron. Meanwhile, the alternative to closed app stores on mobile devices is the "progressive web app." PWAs rely on the ever-accreting blob of web standards to replace native mobile OS functionality. (PWAs are, of course, intentionally hindered by Google and Apple to protect the walled gardens of their app stores.) Meanwhile, web assembly, a technology which allows any program in any language to run in the browser, is poised to become a standard environment for computing across basically all platforms.

Web technologies are now at core of computing, and web browsers are on the hook for just about all of computing. To be sure, web assembly and its system interface (WASI) are actually exciting technologies to me; I'm not casting aspersions. (And my team and I use both WASM and the complexities of the modern web to make our programming language and environment, Ludus easy to use.) We're already using the browser as a replacement for operating systems' graphics toolkits, while simultaneously building something like an operating system inside the browser. (Obligatory, delightful shoutout to this Destroy All Software talk.)

Simplicity and betrayal

Against this complexity, I admit that I'm seduced by the promise of something like the Gemini protocol. Gemini is a new hosting protocol that's both brutally simple and impossible to extend into the mucky ball of complexity that is the modern web. Gemini's main idea is that, for at least a lot of individual users' purposes, the web is way too much; you could radically simplify things and still have a vibrant internet. You could do it without ads. If you do your simplification right, the reasoning goes, the result will support the thriving of those individual users, rather than corporate interests.

I'm very sympathetic to this. I want a web driven by users in control of their technologies: the content and the servers, but also the browsers and the protocol itself. By far, the most important aspect of Gemini lies in the fact that the protocol is simple enough that yours truly could write a lot of a basic (and probably bad) client, by myself, in a weekend or two. And that a small team of dedicated people could maintain an excellent client easily, as an open source project.

If my Mastodon feed was any indication, people felt, and continue to feel, betrayed by Mozilla. (This is, of course, only the most recent disappointment in a long line of disappointments.) This feeling of betrayal is grounded in the fact that we need large, well-funded, deeply compromised organizations to write ghastly complex software that we could not hope to understand or replicate. The depth of that compromise is becoming clearer.

Silly optimists like myself had hoped that Mozilla's nonprofit status might protect it, and us, from the depredations of Big Tech. But it was always funded by Google's payments to remain the default search engine in Firefox, and the antitrust writing is (blessedly) on the wall (although what a Trump 2.0 Department of Justice will do with Google is anybody's guess). Mozilla is, probably correctly, reading the tea leaves and looking to scrounge operating funds once the Google gravy dries up. That's something like "responsible stewardship" if you squint hard enough. Mozilla, as they say, is responding to incentives.

It would be nice to have a better browser, and thus also less damaged ways of using the web. But I get the sense what we need is not an alternative browser, but an alternative to the web. It's not clear what this could be. Gemini inspires longing in me, but these problems are as much matters of political economy as anything else, and I don't see how the current political economy of computing could end up hosting a robust Gemini ecosystem, as much as it inspires nostalgia in me.

There's a plausible position that we ought to just simplify the web, turning to a brutalist motherfucking aesthetic to radically simplify things. I won't even link to complaints about too much Javascript in contemporary web development, so common is that complaint. As nice as that may be (and as actionable as that is on an individual basis: just become a brutalist!), it doesn't fix the widespread problems of the web: browsers are too complicated, websites and web apps hoover up too much of our data, the business models of this particular political economy seem intractably oriented towards Cory Doctorow's enshittification at the platform level, and Ed Zitron's rot economy at the firm level.

Scaling down

Since Mozilla's announcement, and after the huge jump in Bluesky use after Trump's election and Musk's involvement in the new administration, Christine Lemmer-Webber wrote some really, really wonderful blog posts (part 1, part 2) posing and answering the question, "How decentralized is Bluesky, really?" (tl;dr: Not very.)

In those posts, Lemmer-Webber poses one of the most important questions I've seen in a while: how does a technology scale down? In her analysis, this poses this question about the technical complexity, data storage requirements, and general expense of running a second instance of Bluesky using its ATProtocol. And, to get a "credible exit" from Bluesky, she argues that it would be very complex, with huge amounts of storage, and very expensive indeed. Like web browsers, which are mostly open source in any event, getting to an alternative to Bluesky isn't a problem of open standards or protocols, nor even of open source, but of platform scale. The work Lemmer-Webber and her fellow travellers are conducting at Spritely is both impressive and crucial in developing an alternative to centralized computing platforms. (And I love that they're using Scheme.)

But the question of how a computing technology scales down is, to my mind, one of the most important questions of the moment. It is not only a question of, say, the amount of data storage required to host a whole alternative social media platform, as in Lemmer-Webber's analysis. It's also not only a question of how many pages of standards, or how many passing conformance tests, one must satisfy in writing a new browser, as DeVault points out (and Kling rightfully boasts about on his monthly Ladybird progress videos). Scaling down is perhaps, the fundamental question of our current technological political economy. We live in a new gilded age, with monopoly triumphantly resurgent, and billions of people at the whims of the execrable tech billionaire class. Lina Khan isn't going to keep her job, and I live in a country perhaps best described as three telecom monopolies in a trenchcoat. Let's not get started on the newest, biggest, most catastrophic technology Big Tech has cooked up, the Large Language Model and its transformer-based ilk.

Perhaps it's my middle age, and with it the renunciation of youthful aspirations of a big life with big impacts, but I want to find ways of making small, meaningful things for my communities. I want to figure out how to make the biggest impact I can in my own small worlds, or maybe make some new, excellent, likely temporary, small worlds.

There's lots of "small tech" and the "small web" in the world. Perhaps none of it is very visible because it's, of necessity, small. A lot of it--say, 100rabbits' uxn system--gives me the same nostalgia as Gemini. Others are doing other work thinking small, e.g. Aral Balkan and Ben Hoyt. The Spritely folks are doing work that I find extremely interesting on the question of scale. I'm sure there are many others.

My particular talents aren't exactly oriented towards promoting small things, or building usable small systems. My sense is rather that the question of scale, and smallness, in particular, hasn't yet been posed fully or correctly. The critical discourses about computers--media theory, science and technology studies, game studies, and so on--have grown up theorizing a technology surround that was getting bigger all the time. It's time to think well about how to get small.

*top image from https://commons.wikimedia.org/wiki/File:NCSA_Mosaic_Browser_Screenshot.png

Scott Richmond is Director of the Centre for Culture and Technology and Associate Professor of Cinema and Digital Media at the University of Toronto. They work at the intersection of film theory, media theory, experimental media aesthetics, and the history of computing. Their first book, Cinema’s Bodily Illusions: Flying, Floating, and Hallucinating was published in 2016 by the University of Minnesota Press. Their second book, Find Each Other: Networks, Affects, and Other Queer Encounters is under contract with Duke University Press. You can find other writing and information about their current work at https://scottrichmond.me.