Re: Translation conventions

On Thursday 09 April 2009 16:04:32 Greg Stark wrote:
> Hm, I may have made an assumption here which might be wrong. I assumed
> the original English strings were updated regularly whenever we made a
> release or even more often. Even if no translator was available to add
> translations for them. Or do we only regenerate the original English
> strings on demand when someone is ready to translate them?

The flow of things looks like this:

PostgreSQL source code C file
|
| xgettext
|
POT file
|
| msgmerge <--- old PO file
|
merged PO file
|
| human translator
|
updated PO file
|
| cvs commit
|
pgfoundry/pgtranslation CVS repo ---> then serves as next old PO file above
|
| release manager + special scripts
|
PostgreSQL source code with new translations

The xgettext and msgmerge steps are run regularly (several times per day) on
babel.postgresql.org, then the translators do their thing, check the stuff in,
repeat as much as they want, and then before a release usually I do the last
step to copy things over.

To take the example of the outdated backend hr.po translation, if you went to
http://babel.postgresql.org/ you can download the merged PO file, do some
translation, check it back in, and then it will be in a release in the near
future.