A Reflection on the Small Web
misc.l3m.in
misc.l3m.in
I’ve developed a terminal browser for the web/RSS/gemini/gopher and I’m everyday astonished how 95% of what I need is best served as simple text in my terminal. It makes me wonder "why do we to invest so much time learning/tweaking CSS/JS when all we want to do is exchanging text and a few pictures ?"
https://notabug.org/ploum/offpunk
PS: to be inclusive of gemini/gopher, I prefer the word "smolnet" over "small web". But we understand each other, that’s really a detail.
(I read your posts on your gemlog as well as the french ones on your blog, thank you very much for writing all of these posts)
Those "open source" web engines have grotesque and absurd complexity and size, are written with a computer language (c++) which has a syntax complexity morbidely accute that it requires compilers (even naive ones) not better than the web engines themselves.
Hard truth: bazillions of services provided over the net can be provided as noscript/basic (x)html, aka basic html forms.
Hard truth: asking nicely usually does not fix that (I tried).
Hard truth: the real work in not the web "code", it is protecting it against hackers, like DDOS attacks.
Hard truth: Don't fool youself, hackers will be "paid" to push hard those sites on big tech network infrastructure (or are idiots). "Ppl coding computer viruses are probably those who sell anti-virus software".
But now tools such as Google Lighthouse and Next.js logging first load file sizes makes it quite tangible to focus the team on and advocate for faster first meaningful content paints.
That seems like trying to address one of the symptoms, not the underlying problem. "First meaningful paint" means nothing to me, depending on what the developer means by "meaningful". Why can't I just have the whole page load and display quickly?
A lot of this nonsense is because of mobiles. Mobile browsers don't handle some UI elements (checkboxes, radios, drop-downs) very well, and all the browsers have "interesting" incompatibilities; so developers use these compatibility shims, which usually result in a badly-degraded desktop experience, and slowness for everyone.
So I'm blaming the mobile browsers. They don't deal with standard markup properly.
Unless you have something specific you want to communicate and then its up to you to decide whatever you find meaningful to show first.
When I say great content, I mean something that can federate a large audience and serve as a reference and inspiration for others.
Today most of content creator are producing for the rich web, so most of the interaction happen there, and as a result most of the new content is formatted for the rich web.
From the top of my head, Phrack Magazine [1] is the only examples I can think of text based good content. It is mostly aimed at a very specific audience, but I am sure they are other examples out there.
Which basically lead to the other question of discoverability...
You will not find that trendy topic everybody is talking about. You will not find buzz and virality. You will not find a bunch of content creators with thousands of consumers happy to comment/reshare.
And that’s the whole point.
You write in the smolnet thinking you write in the void. Then, sometimes, you realize you are part of a conversation. Because you see posts replying to you. Because you receive emails.
The web is now mostly a huge stadium where we watch millionnaires playing football. The smolnet is the grassy field were you can play with others, having fun and improving your skills.
The small field is not there to replace the huge stadium. You can even enjoy both.
Here is something you should think about. If Small Web actually worked, we would be discussing this on Small Web right now. Do think about what this statement actually means before dismissing it as yet-another-negative-HN-comment.
This means, for starters, that I would gladly use another platform to discuss your ideas, but none of the websites from various "alternative web" movements have any plausible solutions for people collaborating with one another. Which indicates a kind of lack of high-level, systemic thinking that is absolutely essential if you want to change anything at scale.
Well you've pooh-poohed the idea without offering a constructive alternative. To me that's a typical negative HN comment. Why not suggest some ways to make it closer to reality?
Tools like PowerPoint and MS Publisher have allowed (for decades) the ability to just drag and drop anything you want onto a screen, save it, then send it to someone, and it will display for them exactly how it displayed for me. Text, pictures, sound, video. And it supports clickable hyperlinks. Which was the entire point of the web! But for some reason, web browsers are 10,000 times more complicated than these other ancient tools.
If we really want to make the web less braindead, less bloaty, easier to use, we should throw away the annoying crap we have now, and start from scratch with something that is just easy for the average person. We don't have to regress to text terminals and Gopher, or just use browsers without JS. We can make something new that is both easy and useful.
The tools you mentioned have a fixed viewport size, the web does NOT have it.
99% of the difficulties of a web design is in supporting all possibile client screen sizes / ratios.
A powerpoint like experience is trivially to build, just use position: absolute and place everything pixel by pixel.
I'm not aware of any GUI for creating responsive content (unless you limit the content to a simple flow of text)
For a web designer.
For a normal person, 99% of the difficulties of web design is that they need to get a PhD in HTML, CSS, & JavaScript, in order to make some text and images show up. PowerPoint/Publisher are usable by anyone. The Web today is basically an employment scheme.
> I'm not aware of any GUI for creating responsive content
If all you want is to show someone some text and pictures, that should be (is, actually) incredibly simple.
If you want to show someone "responsive content", that is a different use case. It's a custom program at that point. So just ship them a custom program. It used to be common, trivial actually, to mock up a program in a visual IDE and send it to someone. The program would be small and fast and networked. We used a thing called Visual Basic, and nearly anyone with the most basic programming skills could use it. Worked great.
Today we do the same thing, only they are called "mobile apps" and are written in Java for a dedicated mobile operating system. Just like years ago, these are proprietary platforms run by quasi-monopolies, so you have to make your program multiple times to send it to everyone. But we could improve on that.
No it's not! You create your content with text and images, you send it to two friends: one opens it in a 27" inch monitor, the other one on his mobile phone. How big should the images be? How big the text? What happens if there is a text on the left and an images on the right, and you watch it on a mobile phone? should it go down, should the image resize to occupy at maximum half the viewport? what happens when the text overflows?
The reason you can do it on a Oowerpoint or Publisher is that the viewport is the same, you position it and you are done. The user on the desktop will see it very small, the user on the phone will need to keep scrolling and dragging.
If you are ok with these limitations, there are plenty of tools available to share content! You named them.
If you want better control, there are no easy way of doing it, there are no example of this problem being solved, and I don't think it's a web problem.
Here is a screenshot of Android Studio using a point-and-click layout editor to create responsive design, with no manual editing of code: https://developer.android.com/static/images/screens_support/... https://developer.android.com/guide/topics/large-screens/sup... It's very reminiscent of Visual Basic.
Moreover, the layout editor is rarely used in Android development, most of the control is done via code because it depends on dynamic content that is not available to the layout editor ahead of runtime. Most the examples in the exact link you attached are code examples for this very same reason
Not all websites need to be dynamic, though. And I'm with you in spirit. I just think that throwing out JavaScript entirely throws out the baby with the bath-water for a lot of websites.
In theory, you can use XHR and client-side rendering to make things smaller, not bloatier. Fetch the HTML once with first content, then fetch only content updates at the user's request without reloading the page. This can result in better cache efficiency and improved render times. But a lot of people might object on principle to the idea of using XHR requests and dynamic behaviour in something called "small web."
You mean discussing about lightweight websites on a website that weights less than 100KB? That would be incredible for sure :D
>You mean discussing about lightweight websites on a website that weights less than 100KB?
Many HN threads exceed the 50kb download you mentioned in your post. See examples from: https://news.ycombinator.com/best
- over ~200 comments will often exceed 50kb download
- over ~500 comments can exceed 100kb
(Open those big threads in with F12 browser Dev Tools and look at the byte counts.)
The arbitrary byte limits of 50kb or 100kb don't seem like a useful heuristic. In other words, if you tweak the rules to increase it to 200kb to allow HN threads, it then conflicts with other "bad" downloads that seemed to fit the profile of 50kb on other websites being an indicator of something unwanted.
Doesn't HN count as Small Web? It seems pretty minimal.
Not small in traffic.
[0]: Another small rant here: https://misc.l3m.in/txt/html.txt
I'm gonna go out on a limb and say that that is basically what it boils down to - the small web, and similar movements, just aren't for you. Because...
>This Small Web stuff (and other similar movements) is close enough to plausibility to capture people's time and attention, but far enough from reality to inevitably fail without making a dent in the awful state of the Web today.
... in spite of that point, the small web has always been around in one form or another. While the rest of contemporary culture glommed onto Web 2.0 and social media, there were just enough people keeping alive some of the ideals from the older web that have made their way into the more contemporary "small web" movements that you see today. There's always been an audience.
>Which indicates a kind of lack of high-level, systemic thinking that is absolutely essential if you want to change anything at scale.
But I don't see it (and this is just my view of the small web) as an attempt to completely wipe out the current, contemporary, "popular" web. "Scale" isn't the goal, and I'd argue that scaling is antithetical to the general ideal of a small web. It's about focusing on smaller, community-run spaces without all of the noise and bullshit. You have to resist scaling in order to keep that spirit.
To that end, the small web will always be there, for those who want it.
What would failure look like though?
I mean, I have a couple of Gemini sites ("capsules") up on sr.ht, I figure that they will stay available at least as long Drew Devault lives, and I bet he'll take steps to keep the service going beyond himself.
Since I'm not trying to record traffic stats, or monetize, or anything like that, how would failure be defined?
> If Small Web actually worked, we would be discussing this on Small Web right now.
As lambic says in a sib comment, I think HN counts as "Small" (that's a large part of why I use it and not, say, twitter.) Sure it has some JS and CSS, but it's basically text-based and all the dynamism is unobtrusive.
> none of the websites from various "alternative web" movements have any plausible solutions for people collaborating with one another.
Gemini, at least, has facilities for interaction, and some sites make use of them.
Also Gemini itself was developed by people collaborating with one another, largely over email & mailing list. (Interestingly, last I checked the mailing list serve had gone down, and they left it down because Gemini A) is done and B) has developed a self-sustaining community outside of the list.)
(FWIW, I have a mailing list and an IRC channel for collaborating with people, suits me fine. YMMV.)
If you take text file, wrap it with just a <pre></pre> tags, and publish it as .html you can then use any inline html tags such as bold, italics, links, etc. And even block level stuff like headings and lists.
But really, just mark up your text as plain, semantic HTML. It add very little to the file size.
Yes, we want pages small. But we also want images and video in those pages, and those can and will be quite large.
For example, this kind of visual tutorial page is impossible in Gemini: https://woodgears.ca/pantorouter/linear.html
In it's insistence on "human consumption of predominantly written material" Gopher has removed the possibility of such content, too.
Edit: corrected a link
As I mentioned in a comment above, wrap the plain text in <pre></pre> tags, serve as html, then you can use links without adding any other html at all.
echo -e "\xef\xbb\xbf\n" > myfilename.txt
Content-Type: text/plain; charset=utf-8I created the whole https://misc.l3m.in/txt/ trying to prove a point with a friend (the point was something like "you can start a blog right away if you have a domain name and a ssh access to a server, let me show you") :D