l10n approach for bzr

[Mark Hammond]

> class MyError(BzrError):
> 
>   _fmt = i18n("My Error says %(foo)s")
> 
> We would instead do:
> 
> class BzrError(...):
> 
>   def __str__(self):
>     l10nfmt = i18n(self._fmt)
>     return l10nfmt % self.__dict__
> 
> 
> And the alternative:
> 
> class MyError(BzrError):
> 
>   _fmt = i18n("M_BZR_MY_ERROR_FMT")
> 
> Which gets exceptionally tricky if you are changing the 
> number of %s you want to put in the string. You now have to 
> go find the format, figure out that it has 3, add the 4th, etc.

Unless I'm missing something, it is already worse than that - you need to
locate *all* existing translations and add the new insert (else a runtime
error formatting the string will occur in some locales but not others) - and
worse, you need to know *where* in the translated text the insert makes
sense?

> So fairly long winded, but my view is that we should use 
> English strings for the message ids, and that when the locale 
> is English we shouldn't need any translating. Which means 
> that when strings are cleaned up, they will get new message ids.