In other words, it's high time to realize that L10n/I18n is a full needed part of the development cycle, and it does not just consist in translating a few strings in the last few hours before release. It will be already too late to effectively test the correct localization.

Developers in MediaWiki tend to think that this is a minor issue, this is a severe error ! If we need quality translations and better usability for all users, ALL developers should include L10n/I18n in their own cycle (we've got already problems providing workable translations because developers ignored the basic grammatical requirements, or layout requirements for their UI, or because they assumed wrong things about what other languages and scripts really need). 

Not all L10n may be solved immediately before release, but the difficulties requiring further developments should be scheduled so that it will orient correctly further developments. Note that quality translations also require that the new developments are testable somewhere. This is necessay to ensure that the terminologoies will match with the rest of the interface. This includes as well the help/documentation pages to prepare and that will be advertized by the new features.



2013/6/10 Philippe Verdy <verdy_p@wanadoo.fr>
Your short window of 24 hours will never be workable (or it will simply exclude most translators that cannot be tied to this very strict schedule) it is even worse than the 3 days delay we have now (which is in fact already shorter, due to the time needed just to receive the notification : I've already received notifications for announcements that were supposed to occur 3 days later, but whose deadline was already passed since a few hours !).

After all these weekly updates come after weeks (if not months) of developments and testings, and these translations should be announced and prepared long before, during the development, so that these translations would be ready long before the weekly updates announcements.
In other words, we should ask to developers to request translations before, even if this is just for an early beta and that some translations will change several times, but at least the essential would be ready with less to translate for the last phase before launch.



2013/6/10 Guillaume Paumier <gpaumier@wikimedia.org>
It's a difficult choice to make. On the one hand, we would ideally
want as many translations as possible, so a week of lead time would be
better in this regard. On the other hand, most of the content of the
tech newsletter becomes outdated very quickly, and would be mostly
useless if sent a week later.

I'm open to suggestions about how to facilitate translations while
sending information to subscribers before it becomes outdated.

One possibility that I'll try to adhere to as often as possible is to
have a very predictable schedule, meaning for example that the
deadline for having the newsletter ready for translation should be
very strict. That way, translators would know that every week, there
will be a window of 24 hours for translation, and the window will
always be the same each week. I think more predictability would help
facilitate translations.