[takoi]

@takoi: It wouldn't surprise me if it was that high. What would be a huge help is for the contexts to be done for all the strings you found - once the context clearly indicates where it's used, it's both useful to the translator (as you know where to look to see whether things fit etc) and also useful for developers when removing/replacing old strings as they know they are unused now. You can fill in something like "unused" for the ones that aren't done and I'm sure that Jezz_X will be happy to take a hunt through and confirm that they're unused, thus removing them.

Obviously this is a large job, but if everyone pitches in and just does a few each (just let us know which block you're doing), it won't take long before the whole lot are done.

I am planning to write a parser utility to extract at least the occurence of the string calls. The tricky thing is that we have several ways of passing the string id to the localization function. Sometimes it comes from a direct parameter, but sometimes it comes from an xml file or from other indirect calls.
So whenever I have time for it, I'll try do it.

This is a perfect example of why string keys are so important.

If everything is moved to strings (wrapped with direct calls) you have essentially moved from maintaining a .pot file to generating. Like gettext does. The unused string problem is solved. The extract occurrences problem is solved.

Maybe a temporary solution (before the id to string move is done) could be a simple file storing all the id to string mappings. Then, the English .po template is generated by extracting the IDs from source code and looking up the string in that file (since you're planing to write that parser tool anyway)?