Hello,

if I understood everything right the procedure at the moment is the following. 
Please correct me if I am wrong.

Situation 1:
A program is already translated upstream and there is no corresponding 
translation in Rosetta. --> The upstream translation gets imported.

Situation 2:
There is a new version of a program, that is to be imported to Rosetta. It is 
translated upstream but there are changes since the last version imported to 
Rosetta. The old "upstream translation" in Rosetta has never been changed by 
translators in Rosetta. --> The new upstream changes get imported 
automatically.

Situation 3:
There is a new version of a program, that i to be imported to Rosetta. When 
the last release was imported, there had not been a complete upstream 
translation, but the missing strings have been translated by Ubuntu's 
translators in Rosetta. Now the new version has a complete upstream 
translation. --> The new upstream changes get imported as suggestions. The 
existing translation in Rosetta stays as it is. Rosetta takes priority over 
upstream.

Situation 4:
There is a new version of a program, that is to be imported to Rosetta. When 
the last release was imported, there had not been a complete upstream 
translation, but the missing strings have been imported from upstream's svn, 
where they already had been translated. Now the new version has a complete 
upstream translation with some string changes. --> The new upstream changes 
get imported as suggestions, because Rosetta considers the existing 
translations as done by Ubuntu's translators. These translation in Rosetta 
stays as it is. Rosetta takes priority over upstream.

Situation 5:
A program or a new version of it is to be imported to Rosetta. But the import 
fails. So there is an untranslated or only in parts translated template in 
Rosetta. The Ubuntu translation teams start to translate this. Now someone 
notices the problem and the upstream translation gets imported. Actually this 
happens for example with ktorrent [1]. --> The now correctly imported 
upstream translation gets imported as suggestions. The existing translation 
in Rosetta stays as it is. Rosetta takes priority over upstream.

Situation 6:
There is a new version of a program, that should be imported to Rosetta. The 
last release was imported completly but some strings have been changed by 
Ubuntu's translators. The same strings have been changed upstream as 
well. --> The new upstream changes get imported as suggestions. The existing 
translations in Rosetta stays at it is. Rosetta take priority over upstream.

I think there is no need for discussion about situations 1, 2 and 6. Of cause 
it is right to keep changes manually done by Ubuntu's translators, because 
they will only edit a existing translation if the new translation fits 
better to Ubuntu. The next release it is possible to see if the new upstream 
translation is better, but probably the reasons for changing it stay the 
same.

Situations 3, 4 and 5 are quite different in my eyes. If the translation in 
Rosetta takes priority over the upstream translation the result would be a 
fork, because Ubuntu don't recieve changes from upstream automatically and 
has its "own translation", different from all other distributions. At least 
if no one changes the translations to the suggested upstream translations 
manually. But this will probably never happen in the majority of teams, just 
because of the lack of manpower. Especially in situation 4 this is 
problematic, because the translation is not really Ubuntu's own translation 
but an outdated upstream translation.

The idea behind Rosetta is (correct me if I am wrong) to close the gap between 
distribution users and upstream translation teams, to add missing 
translations and to improve the existing translations or adapt it to the 
needs of Ubuntu. I think to achieve this goal it is better to change the 
behaviour of Rosetta in situations 3, 4 and 5 in the following way: Import the 
upstream translation automatically and add the Rosetta one as a suggestion.

This would probably reduce the rejection of Rosetta by the upstream 
translation teams noticeably. And it would be still possible to keep an 
existing (and perhaps better) Ubuntu translation, but the risk of forking the 
translation accidentally is banned. In all the cases it is resonable to take 
priority over the upstream translation this would still happen, because this 
would only affect "new translated" but not "changed" strings.

Please think about it.

Kind regards,

Jannick Kuhr
(member of the german KDE translation team and of the german Ubuntu 
translation team)

P.S. I sent these message as copy to kde-i18n-doc@kde.org

[1] https://launchpad.net/bugs/58168