I think everyone will agree that Rosetta’s interface is completely unhelpful as it’s slow (I’ll have a drink for each timeout error), cumbersome to navigate and generally badly designed.
The Ubuntu-make project already uses the SAAS version of Weblate, so I know some people at Canonical are already familiar with it.
[1] By the way: Fedora’s recently abandoned their own solution, Zanata, which had been languishing for years due to scarce development resources, and migrated to this. It’s an excellent example of beating the NIH syndrome.
I haven’t seen the timeouts in Launchpad Rosetta, are they reported as bugs?
I appreciate it may be frustrating to use, but we’re 1 month away from release of 20.04, and I don’t believe now is an ideal time to make such a fundamental change. Perhaps something to investigate post-release, for sure.
@popey Never in my post did I say that we must do this now.
@gunnarhj Thanks for the cross-ref, I knew this was discussed years ago. It is disappointing that nothing has come out of that though.
To be clear, since I joined Ubuntu in 2010, translators have voiced their grievances in dealing with Launchpad as a l10n platform, in many places, and since Launchpad’s development is slower than the planet is heating, and given that most Ubuntu development has been modernized to use Git repositories, it just seems more practical to use an open-source, self-hostable solution that is easier to use, which integrates directly with Git and that is in active development.
From https://josm.openstreetmap.de/ticket/8645, four months ago: “Lately it takes 5 to 10 reloads at Launchpad to go to the untranslated strings due to the infamous Timeout error.”
And the cherry on top is Canonical’s own list of basic improvements (under “Ideas”) that other platforms implement but are missing in Launchpad: https://trello.com/b/6sxl9bez/translations-team
In fact, I tried to contact Rossetta developers addressing all the issues with no luck. I’ve been waiting 2 years for a reply.
The platform it’s programed using a deprecated library from Yahoo.
According to some recent rants by the community (Which can be settled with configuring the ACL) I thought that if Canonical would like to invest in its own Weblate instance it could be a very good starting point for silent transformation while preserving most of the requirements.
