Hi Michal, Thanks for the lightning-fast reply!
On 20/03/2020 14:00, Michal Čihař via Weblate wrote: > I know projects translating Sphinx or Publican based docs in Weblate: > > * https://docs.godotengine.org/en/stable/ > * https://docs.phpmyadmin.net/en/latest/ > * https://debian-handbook.info/ > > Fedora is using some more complex tooling for the docs, see > https://fedoraproject.org/wiki/User:Jibecfed/fedoradoc-antora-localization Those pointers are super useful, thanks! > >> We have identifier Docusaurus [2] as a potential good candidate to >> generate our docs, which comes with an integration with Crowdin for >> translation. I wonder if anyone has tried to use Weblate with >> Docusaurus? Interoperability with other translation systems seems to >> be >> on their roadmap for their v2 [3], but it does not seem to be usable >> yet, so I wonder if anyone has experience with setting that up. > > No experience with that, do you know how it stores the translations? I > hope they did choose some standard format for that and in that case it > would be easy to use with Weblate :-). > Crowdin seems to have a very different architecture: translations are not stored on the repository being translated. Instead, translations are stored in Crowdin itself, and one needs to retrieve the translated files at build time. For instance, say you have a Jekyll site with a bunch of Markdown files that you want to translate, you would change the build process (the step where you invoke Jekyll) to also: - push your markdown files to Crowdin (letting it know what has been updated and should be translated); - pull the translations frow Crowdin, returned as Markdown files themselves; - let Jekyll build the site as usual (it does not need to be aware that some of these files are translations of others). This CircleCI workflow seems to summarize the process pretty well: https://docusaurus.io/docs/en/translation#automated-file-sync-using-circleci With this architecture, I am obviously concerned about vendor lock-in (especially since Crowdin is not FOSS), lack of versioning and keeping track of translation authorship in the version control system. If we wanted to use Weblate in place of Crowdin, it seems to me that we would need to introduce a piece of software in between: something that takes markdown files, breaks them down into translatable units which would be stored in some standard format (say a POT for gettext), and be then able to build translated versions of these markdown files using the gettext PO files (and the original markdown). For instance: https://po4a.org/man/man7/po4a.7.php https://www.npmjs.com/package/gettext-markdown Would you recommend any of these? Best, Antonin _______________________________________________ Weblate mailing list -- weblate@lists.cihar.com To unsubscribe send an email to weblate-le...@lists.cihar.com