Elmord's Magic Valley

Software, lingüística e rock'n'roll. Às vezes em Português, sometimes in English.

Some thoughts on Gemini and the modern Web

2022-02-20 19:22 +0000. Tags: comp, web, ramble, in-english

The web! The web is too much.

Gemini is a lightweight protocol for hypertext navigation. According to its homepage:

Gemini is a new internet protocol which:

  • Is heavier than gopher
  • Is lighter than the web
  • Will not replace either
  • Strives for maximum power to weight ratio
  • Takes user privacy very seriously

If you’re not familiar with Gopher, a very rough approximation would be navigating the Web using a text-mode browser such as lynx or w3m. A closer approximation would be GNU Info pages (either via the info utility or from within Emacs), or the Vim documentation: plain-text files with very light formatting, interspersed with lists of links to other files. Gemini is essentially that, except the files are remote.

Gemini is not much more than that. According to the project FAQ, “Gemini is a ‘less is more’ reaction against web browsers and servers becoming too complicated and too powerful.” It uses a dead-simple markup language called Gemtext that is much simpler than Markdown. It has no styling or fancy formatting, no client-side scripting, no cookies or anything that could be used for tracking clients. It is designed to be deliberately hard to extend in the future, to avoid such features from ever being introduced. For an enthusiastic overview of what it is about, you can check this summary by Drew DeVault.

I’m not that enthusiastic about it, but I can definitely see the appeal. I sometimes use w3m-mode in Emacs, and it can be a soothing experience to navigate web pages without all the clutter and distraction that usually comes with it. This is big enough of an issue that the major browsers implement a “reader mode” which attempt to eliminate all the clutter that accompanies your typical webpage. But reader mode does not work for all webpages, and won’t protect you from the ubiquitous surveillance that is the modern web.

I do most of my browsing on Firefox with uBlock Origin and NoScript on. Whenever I end up using a browser without those addons (e.g., because it’s a fresh installation, or because I’m using someone else’s computer), I’m horrified by the experience. uBlock is pretty much mandatory to be able to browse the web comfortably. Every once in a while I disable NoScript because I get tired of websites breaking and having to enable each JavaScript domain manually; I usually regret it within five minutes. I have a GreaseMonkey script to remove fixed navbars that websites insist on adding. Overall, the web as it exists today is quite user-inimical; addons like uBlock and NoScript are tools to make it behave a little more in your favor rather than against you. Even text-mode browsers like w3m send cookies and the Referer header by default, although you can configure it not to. Rather than finding and blocking each attack vector, a different approach would be to design a platform where such behaviors are not even possible. Gemini is such a platform.

The modern web is also a nightmare from an implementor side. The sheer size of modern web standards (which keep growing every year) mean that it’s pretty much out of question for a single person or a small team to write and maintain a new browser engine supporting modern standards from scratch. It would take years to do so, and by that time the standards would have already grown. This has practical consequences for users: it means the existing players face less competition. Consider that even Microsoft has given up on maintaining its own browser engine, basing modern versions of Edge on Chromium instead. Currently, three browser engines cover almost all of the browser market share: Blink (used by Chrome, Edge and others); WebKit (used by Safari, and keeping a reasonable portion of the market by virtue of being the only browser engine allowed by Apple in the iOS app store); and Gecko (used by Firefox, and the only one here that can claim to be a community-oriented project). All of these are open-source, but in practice forking any of them and keeping the fork up-to-date is a huge task, especially if you are forking because you don’t like the direction the mainstream project is going, so divergencies will accumulate. This has consequences, in that the biggest players can push the web in whatever direction they see fit. The case of Chrome is particularly problematic because it is maintained by Google, a company whose main source of revenue comes from targeted ads, and therefore has a vested interest in making surveillance possible; for instance, whereas Firefox and Safari have moved to blocking third-party cookies by default, Chrome doesn’t, and Google is researching alternatives to third-party cookies that still allow getting information about users (first with FLoC, now with Topics), whereas what users want is not to be tracked at all. The more the browser market share is concentrated in the hands of a few players, the more leeway those players have in pushing whatever their interests are at the expense of users. By contrast, Gemini is so simple one could write a simplistic but feature-complete browser for it in a couple of days.

Gemini is also appealing from the perspective of someone authoring a personal website. The format is so simple that there is not much ceremony in creating a new page; you can just open up a text editor and start writing text. There is no real need for a content management system or a static page generator if you don’t want to use one. Of course you can also keep static HTML pages manually, but there is still some ceremony in writing some HTML boilerplate, prefixing your paragraphs with <p> and so on. If you want your HTML pages to be readable in mobile clients, you also need to add at least the viewport meta tag so your page does not render in microscopic size. There is also an implicit expectation that a webpage should look ‘fancy’ and that you should add at least some styling to pages. In Gemini, there isn’t much style that can be controlled by authors, so you can focus on writing the content instead. This may sound limiting, but consider that most people nowadays write up their stuff in social media platforms that also don’t give users the possibility of fancy formatting, but rather handle the styling and presentation for them.

* * *

As much as I like this idea of a bare-bones, back-to-the-basics web, there is one piece of the (not so) modern Web that I miss in Gemini: the form. Now that may seem unexpected considering I have just extolled Gemini’s simplicity, and forms deviate quite a bit from the “bunch of simple hyperlinked text pages” paradigm. But forms do something quite interesting: they enable the Web to be read-write. Project such as collaborative wikis, forums, or blogs with comment sections require some way of allowing users to send data (other than just URLs) to the server, and forms are a quite convenient and flexible way to do that. Gemini does have a limited form of interactivity: the server may respond to a request with an “INPUT” response code which tells the user browser to prompt for a line of input, and then repeat the request with the user input appended as the query string in the URL (sort of like a HTTP GET request). This is meant to allow implementing pages such as search engines which prompt for a query to search, but you can only ask for a single line of input at a time this way, which feels like a quite arbitrary limitation. Forms allow an arbitrary number of fields to be inputted, and even arbitrary text via the <textarea> element, making them much more general-purpose.

Of course, this goes against the goal stated in the Gemini FAQ that “A basic but usable (not ultra-spartan) client should fit comfortably within 50 or so lines of code in a modern high-level language. Certainly not more than 100.” It also may open a can of worms, in that once you want to have forums, wikis or other pages that require some form of login, you will probably want some way to keep session state across pages, and then we need some form of cookies. Gemini actually already has the idea of client-side certificates which can be used for maintaining a session, so maybe that’s not really a problem.

As a side note, cookies (as in pieces of session state maintained by the client) don’t have to be bad, the web just happens to have a pretty problematic implementation of that idea, amplified by the fact that webpages can embed resources from third-party domains (such as images, iframes, scripts, etc.) that get loaded automatically when the page is loaded, and the requests to obtain those resources can carry cookies (third-party cookies). Blocking third-party cookies goes a long way to avoid this. Blocking all third-party resources from being loaded by default would be even better. Webpages would have to be designed differently to make this work, but honestly, it would probably be for the best. Alas, this ship has already sailed for the Web.

* * *

There are other interesting things I would like to comment on regarding Gemini and the Web, but this blog post has been lying around unfinished for weeks already, and I’m too tired to finish it at the moment, so that’s all for now, folks.

Comentários / Comments (3)

Anon, 2022-11-23 13:55:43 +0000 #

> whereas what users want is not to be tracked at all

I think we should not mix concepts. What the user wants, what the user needs and what is ethical for your provider to do are three different ideas, even though sometimes they intersect.
What most of them really want (and market effects and their actual behavior seem to agree with this) is a responsive web that provides entertainment and has as few annoyances as possible (this means few or no intrusive ads mostly).
Most users don't really care about tracking, they are mostly unaware of it. And indeed they are quite happy that ads are at least discrete and relevant (i.e. targeted) compared to the olden days, so one may argue that, due to a lack of dimension of how problematic this tracking can be, they are actually _glad_ to be tracked.
Then there's what's ethical. Of course, the fact this tracking is excessive and used to influence people both into consumerism and particular opinions makes it unethical. But this has nothing to do with what users _want_.


Anon, 2022-11-23 14:05:09 +0000 #

As an addendum
I didn't think of writing why I was being pedantic about it. The distinction matters because that's the predictor of why all these kind of efforts, as noble as they are, are doomed to fail: you will always need funding to have stuff that is interactive, abundant and always online (which is what the users want), and that means either users paying directly (again, for that you need to give them what they want, not what you should) or indirectly (advertisers paying for it from money they get from people buying their products thanks to ads, which are much more efficient when targeted). Users are also quite unlikely to go for giving money directly. The privacy-respecting web is simply economically not viable at world scale.
I also never went into what the users _need_. This is in part because that's much more complicated to find out, but two things we can be sure about is that they need to avoid the tracking and that they need access to the content, which as a corollary means it needs to still be trivially easy to use and cheap or free (of charge). It needs also forms and sessions and what not, since the web already made working and keeping up with news and with some civic duties depend on the web and I don't think there's going back (tho the latter should definitely not make the web mandatory!). But other than that, there's little way to know for sure what the needs are.


Vítor De Araújo, 2022-11-23 15:25:16 +0000 #

@Anon: Those are excellent points, thank you for your comment.


Deixe um comentário / Leave a comment

Main menu

Posts recentes

Comentários recentes

Tags

em-portugues (213) comp (144) prog (70) in-english (58) life (48) pldesign (36) unix (36) lang (32) random (28) about (28) mind (26) lisp (24) mundane (22) fenius (21) web (20) ramble (18) img (13) hel (12) rant (12) scheme (10) privacy (10) freedom (8) copyright (7) academia (7) music (7) lash (7) esperanto (7) bash (7) shell (6) home (6) mestrado (6) android (5) misc (5) emacs (5) conlang (5) politics (4) latex (4) etymology (4) book (4) editor (4) worldly (4) php (4) kbd (3) security (3) tour-de-scheme (3) c (3) film (3) wrong (3) network (3) cook (2) poem (2) wm (2) treta (2) comic (2) philosophy (2) lows (2) llvm (2) physics (2) golang (1) perl (1) old-chinese (1) x11 (1) en-esperanto (1) audio (1) translation (1) kindle (1) german (1) pointless (1)

Elsewhere

Quod vide


Copyright © 2010-2023 Vítor De Araújo
O conteúdo deste blog, a menos que de outra forma especificado, pode ser utilizado segundo os termos da licença Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International.

Powered by Blognir.