A Minsky machine in ncurses terminfo
seriot.ch
ncurses terminfo can simulate a two-counter Minsky machine, making repeated parameter expansion computationally universal
seriot.ch
These days though, terminals are all software. It would be very easy to just ask them. Instead we still insist on shipping a fixed terminfo or termcaps DB. As a consequence, every new terminal emulator is not in there, and so in order to give their users a tolerable experience, they all lie and say they're xterm-256color. Which completely removes any incentive to actually add them to terminfo/termcaps.
So we have this hellishly complex, nearly Turing complete compatibility system that does no useful work at all anymore.
https://github.com/yonasBSD/foot#xtgettcap https://github.com/jquast/ttyscan#motive https://ucs-detect.readthedocs.io/results.html#terminal-capa...
Isn’t the terminal info easily modifyable? You can add new entries at both the system and user level.
The only actual issues are capabilities. But just like all protocols, there’s need to be an agreement between the software and the terminal. And gradual degradation if some capability is not there. I have more issue with some software hardcoding escape sequences than with not being able to add new terminal entries.
And no, caniuse or browserlist are not that. Yes they contain data on what browsers support what, but that's only to guide your compatibility decisions. You can still opportunistically try to use a feature and a new browser that's not in the DB can handle it correctly without having to lie in its UA string.
Do remember that both the interface (the terminal) and the software environment is under the control of the same user. And terminfo is more of a translation layer than a database, more like the Linux virtual filesystem which hides away the different filesystem implementations. Or ALSA and DRM.
Going back to the issue on the software side, many time, the only check they do is checking if stdout is a file or a tty. If it’s the latter, their output is full of color and spinners, no matter if TERM is explicitly ’dumb’.
Querying terminals, in the 1970s when there was a Cambrian explosion in terminal capabilities, with tens of different manufacturers?
Even if you could have gotten all of them to agree this was necessary, I think nobody could afford to put the smartness for such a protocol in their hardware. Hardware was expensive in those days.
In the era of LLMs articles such as these do nothing for me any more.
This article is mostly, if not entirely, LLM-generated. It’s done in a slightly less blatant way than most, but the tells are there, and pangram agrees.
But even if the article was human-written, I’d still not find it impressive now. There’s no way of telling a human was behind the discovery.