Mallard and Ubuntu-Translators and Translation Regressions

[Kyle Nitzsche]

    "A billion dollars here, a billion dollars there, and pretty soon 
you are talking real money!"

I would like to get a subject on the record that deserve explicit 
consideration respecting the proposed transition to mallard:
 * Work for ubuntu-translators
 * Likelihood of lower translation completeness for docs for some time

Taken together, I think these amount to another argument in favor of 
transitioning to mallard in increments, rather than all at once.

Consider that:
 * When a string changes, it has to be translated
 * Into something like 50 languages
 * And sometimes that translation has to be reviewed and accepted by 
another person
 * The transition to mallard is also a translation to topic-oriented 
content, which will involve some re-writing
 * Thus, there will likely be  some, perhaps many, string changes for 
each topic, multiplied by the 50 or so languages.

That's a lot of new translation work.
 * Has the actual quantity of this been estimated?
 * Have discussions been held with Ubuntu-Translators to see how this 
estimated work fits in with their workload?

Bear in mind that some language teams are not as active as others, thus 
there might reasonably be expected to be regressions in translation 
completeness.

This probable regression is a cost to doing this. It may be a cost worth 
carrying, but it is one that has consequences. For example, in almost 
every OEM project, the client writes bugs to the effect that help 
translation is incomplete. We have to explain that the community is 
doing the best it can. While not a critical bug, this definitely affects 
the quality of the deliverable and of the user experience.

These issues (work for U-T and translation regression) could be 
mitigated by implementing mallard on a topic-based schedule spread out 
over a few releases. (This approach would have other benefits I've 
discussed elsewhere.)

(ccing David Planella the translation community liaison.)

Cheers,
Kyle