The process between the TP and Weblate is quite different. As noted, the TP has a dedicated team, smaller and possibly higher quality.
However, remember there was a lot of resistance to host the manual, and we never got the packages hosted at the TP I believe (or was it the other way around?). The release cycle is different. With Weblate, we can translate strings as they come, whereas on the TP, it would be an intense process for translators, right before the release. If a translator misses the deadline, it won't get merged before the following release one year later, which I think is a pretty bad experience. In addition, the process of updating the pot files on the platform is completely automated with weblate, but requires a bit of human intervention for the TP. I believe the workload is similar for merging translations from the platform. Even if we cannot enforce a policy on translators, we can self-impose (as committers or team members?) not using machine-generated content to fill or review translations. It's a bit weak compared to the rest of the pledge but it's better than nothing at all. Even for the TP, I don't think we can force translators to make a pledge about anything. Le 19 juin 2026 13:23:12 GMT+02:00, "pelzflorian (Florian Pelz)" <[email protected]> a écrit : >Hello all. In the context of Guix’ proposed pledge against generative AI: > https://codeberg.org/guix/guix-consensus-documents/pulls/13/files > https://lists.gnu.org/archive/html/guix-devel/2026-06/msg00075.html > >Ludovic Courtès <[email protected]> writes: > >> Hello! >> >> "pelzflorian (Florian Pelz)" <[email protected]> skribis: >> >>> To me, there’s only one realistic way not to make a false promise with >>> this pledge: Stick with taking translations from the Codeberg Weblate, >>> but don’t mention translations nor Weblate in the pledge. Generative >>> LLM AI are not really the problem in translations, though we can >>> discourage its use in a possibly ignored Weblate message. >> >> OK, we can do that. > >Thanks; yes we should if we stay on Codeberg Weblate. > > >>> Or less realism: Set up our own Weblate for Guix, with few translators >>> who maybe better know Guix (which I will not set up; this likely is not >>> a real option in the near future). >> >> From your comments, I feel like leaving the Translation Project was >> perhaps not such a good idea? In the end, the TP is structured around >> language team; it is “small and slow”, but probably higher-quality? >> >> Thanks, >> Ludo’. > >Hmm the Translation Project. https://translationproject.org/ > >Certainly it is a better choice than hosting our own Weblate. If we do >want to be strict about LLM translations, I strongly believe TP is the >best choice. As Ludo writes, TP in my opinion indeed is “small and >slow, but probably higher-quality”. > >If we are *not* strict, TP is a mixed blessing. I believe TP gives a >high quality but low quantity of translations, meaning it is good for >guix core and manual, bad (?) for packages. The website is basically >done anyway currently; Weblate had been good for its simple topics. > >Yes, I would favor leaving Weblate instances and going back to the TP, >though it makes it harder to compare translations among languages. But >TP has copyright headers. No screenshots though and no automation for >French interpunctuation. > >Do we all agree we want to move? > >Cc other frequent Weblate translators Andrei Șova, F. Serrador, N4ta if >they want to give input. And most importantly Julien Lepiller who >proposed Weblate; it was right at that time. > >I cannot contact Codeberg translator Umeaman, kostic, gallegonovato by >mail. > >Regards, >Florian
