Lokalization for KDE AppStream AppData files

[Albert Astals Cid]

El Dijous, 20 de febrer de 2014, a les 19:33:50, Martin Graesslin va escriure:
> On Thursday 20 February 2014 18:54:32 Matthias Klumpp wrote:
> > Hi!
> > I am working on bringing AppData to KDE. AppData is a XML-based
> > metadata format to enhance information displayed about applications in
> > software-centers. It, for example, includes long descriptions of an
> > application, homepage links, donation-links, screenshot-info and some
> > other information about the application.
> > AppStream is a Freedesktop project enabling us to build cross-distro
> > software centers. GNOME is already making heavy use of it for their
> > GNOME-Software application, and KDE also uses it already in Apper and
> > probably Muon soon.
> > One challenge for KDE is how to handle localizations for AppData
> > 
> > information. I would propose the following:
> >  * KDE applications ship an <app>.appdata.xml.in file, containing the
> > 
> > raw, untranslated data.
> > 
> >  * We extend messages.sh to scan the XML file and return the found
> > 
> > strings to the generic pot file, which is then translated by our
> > localization teams as usual.
> > 
> >  * At build-time, we use intltool[1] to merge the translation with the
> > 
> > AppData file, then install it into it's target directory.
> > This is basically it. The only thing KDE projects need to do is to
> > edit their Messages.sh and depend on intltool to merge localization. I
> > could also provide a cmake macro for doing that to
> > extra-cmake-modules, if needed.
> > Do you have comments or suggestions on this?
> > I would later add it to the wiki page I'm doing to describe how KDE
> > projects which want to use AppData can easily use it.
> 
> Why can't scripty update the xml file like it does with the .desktop files?

This is how has to be done.

Cheers,
  Albert

> I'm a little bit afraid of a more complicated build process as the
> translations life in a different repository and I assume most devs don't
> have them around at all.
> 
> Would that work for distributions? Don't they build the language data
> separately?
> 
> Cheers
> Martin