I previously proposed adding a subrepository for translating documentation to Rust, but it was closed due to lack of capacity to maintain it on an ongoing basis.
So, my other idea is to set up a GitHub organization (e.g. rust-lang-translations) and domain (e.g. rust-lang-translations.org) for translations, and centralize all documentation translations there.
If this proves that there are sufficient translation resources, we can use that as a basis to consider merging it back into Rust itself,
but if no one joins, we can conclude that there was no demand for translations in the first place.
If you have any thoughts on this idea, please let me know.
Personally, with observing various efforts on i18n areas going on, frankly i'm a little doubtful about whether volunteer based translation can serve well in the long run. Personally I feel this area might be better tracked by Rust Foundation, outside the rust Project itself,
Process
With the recent progress of LLM-based technology though, i've been wondering if it's possible to set up a cheap, deterministic, auditable(like telling it to use well-defined translated terms that's conventional to the community, or teaching it to express certain concepts better), automatic process that generated some baseline translation artifacts of Rust documents. I know many software companies like microsoft has been serving such machine-translated materials for their technical documents, but i don't really know much about the technical details inside.
I agree that volunteer-based translation doesn't last long. In my experience managing Japanese translations, the task of extracting updates from the original text is particularly burdensome, and the prospect of fully automating this task with mdbook-i18n-helpers was one of the reasons for starting this project.
Of course, it would be ideal to have support from the Rust Foundation, but there are probably many other high-priority tasks, and even if there is support, there will be issues such as which languages to target, so I think we have no choice but to make it volunteer-based.
I also have high hopes for LLM. One of the goals of this project is to deliver workflow improvements to all documents and all translators by aggregating workspaces. I think the LLM-based initial translation generation function fits that goal.
By the way, although it wasn't announced on this forum, the project has already started at the following link, and at least some of the Japanese translations are scheduled to be merged here.
Firefox (and some other browsers as well) already offers machine translations for web pages. Is there a use for maintaining and hosting official machine translations, when similar quality is already available for almost everyone at convenience?
I know several people who have university degrees in translation. It's not a trivial art at all, and writing a good translation can be surprisingly involved. Maintaining translations written by professionals is something I would wholeheartedly support, but I'm very critical at spending resources on machine translations, especially considering their (almost) trivial availability.
I think the usefulness of machine translation probably varies depending on the language. For example, in Japanese, it is not uncommon for completely incomprehensible strings to be generated, or for the meaning to be completely reversed. (I wonder if it is more accurate in European languages?)
It is useful for skimming through GitHub issues and forum threads, but I get the impression that it is not accurate enough for beginners to use when learning new concepts such as lifetimes.
Therefore, in this project, we believe that users who actually need translations can translate them in their own language communities. If no users want to translate, that is proof that machine translation is sufficient, and it is time to end the project.