ML @ml@qoto.org

@tzerenij Strange, I wonder what happened. The only explanation I can think of is that I hadn't switched from melpa.milkbox.net to melpa.org. I only discovered that melpa.milkbox.net wasn't working anymore after I had already upgraded to emacs 27, so maybe what actually fixed it was just fixing my melpa

Learned about Gemini today from various fediverse users mentioning it in their profiles. I wanted to use the emacs client elpher to check it out, but it required emacs version 26.2 at the minimum, but my version from Debian's repos was 26.1, so I built the latest version of emacs from source and fixed a few incompatibilities between my init file and version 27, and now I'm in!

@fikran I would have agreed with you entirely about the GPL in the past, though lately I've heard about interesting new licenses. The Parity Public License in particular is interesting to me, since it goes even farther than the GPL by even requiring software you "develop, operate, or analyze" with it to also be open-sourced.

(Though note that the OSI does not consider Parity to be open-source because this requirement is too strong and they consider that it impinges on the right to run the software for any purpose).

https://paritylicense.com/versions/7.0.0

I admit that I might be going too far in the copyleft direction by liking the Parity license, but hopefully putting a 10-year limit before it enters the public domain mitigates my overzealousness.

@fikran I like copyleft, I think it has its place in keeping software free. But copyleft is kind of a legal hack, and you get competing incompatible versions of it (like GPLv2 vs GPLv3), so it's better if it's not quasi-permanent like it is today.

Gotta mute a lot of bots to use the federated timeline, that's kind of unfortunate. Or maybe there's a way to filter them? Hmm

Something which I haven't heard about, but which I think is a good idea, is to license free software projects under a strong copyleft license, but add a clause so that it becomes public domain after X years (my preference is 10).

This would have a few advantages. One advantage is that it would be more consistent with my desire to reduce copyright term lengths.

Another advantage is that it would make re-licensing easier, while still being difficult enough that it couldn't be done on a whim. In the worst case everyone could dual-license every new contribution under both the old and new licenses for 10 years, and then 10 years later it would be under the new license. This is just the worst case, though. If you got every contributor for the past Y years to agree to re-license, then you'd only need to wait for 10 minus Y years.

A broad consensus would still be required to re-license the project, so it would still generally be safe from malicious re-licensing, but it would also be easier to fix licensing mistakes (for example, if the project was GPLv2 but the community wanted to switch to GPLv2+). It would also become less risky to experiment with different licenses, because it would be easier to correct the mistake if it became clear it *was* a mistake in the future.

This greater allowance for experimentation would also be useful, I think, because there exist interesting new copyleft licenses which might be a good idea or which might not, like the Parity license.

Technically I've already applied this idea today in some small things I uploaded to github, but they're not projects used by anyone else, so I'm not sure they really count.

Anyway, if anyone is curious, here's a link to the contents of a "LICENSE.txt" file applying this idea, for a project whose last contribution date was in 2019. The idea is that you'd bump up the year in this file each year, so older versions would gradually enter the public domain 10 years after their release. Note that I'm not a lawyer, so I can't actually vouch for this being legally valid.

https://raw.githubusercontent.com/ml-2/sslib/master/LICENSE.txt

The rules of this instance are very good. I like that there's LaTeX, as well as a large character limit.