@dillo If we drop scripting (including css) and the expectation of the same rendering in all user agents (from what people expect user agents to do), it may make writing new user agents easy enough to have an appreciable effect, even without simplifying the HTML grammar a lot. And new user agents should support old websites written in non-XML HTML—reimplementations of Web-like things exist (for example gemini), but after all, I want a fork, not something new.

Does something exist or can we establish something like an open “small user agents group” or whatever where we can start by collectively writing on a website “Let it be known: User agents are henceforth not expected to implement CSS or JavaScript”?

@SRAZKVT but lex(1) to the rescue! you can have a line that matches aaabc and a line that matches aaacd and they get their own actions and it’s still compiled to a single automaton and the aaa will still only have to be matched once

lex is so cool

@SRAZKVT but that’s slooow

if you want to match /aaabc/ and /aaade/, matching /aaabc|aaade/ won’t have to match the /aaa/ twice

@SRAZKVT @fiore

@navi said a little about semantics; “no links” I can refute: groff can convert references to other man pages (.MR/.Xr) and references to URIs (.UR/.Lk) to hyperlinks on terminals (via OSC 8), in HTML (<a/>), and in PDFs. as an example for that last one, maybe take a quick look at this PDF version of the skarnet-man-pages I mentioned in my other reply: https://ljabl.com/tmp/ska-man.pdf (very much work in progress)