We need more solid guidelines for i18n in OSS projects

Every time I'm about to tackle i18n in some project, I find myself filled with a bit of trepidation. I actually know more than the average developer on the topic, but it's a hard topic, and there is quite the mystery around it.

One of the reasons for this mystery is that there are really few good resources to guide FOSS developers through the process of internationalizing their apps. Gettext is the most common means for i18n and l10n in FOSS apps. The master resource for this is the GNU gettext manual, but that's a lot to bite off. Poking around on Google, I found a useful gettext overview for PHP folks, the Ruby on Rails folks have a nice chapter on the it and there are a few other general intros (1, 2, 3). Luis Miguel Morillas pointed out some for GNOME, KDE and wxPython (1, 2).

Python has the usual set of gettext-based facilities, which is great, but they are woefully documented. In the library reference's section on gettext you get a purported overview and a bunch of scattered notes on the API, but nothing that really coherently leads the developer through the concepts and process of i18n, as the PHP and Ruby folks seem to have. It doesn't even seem as if buying a book helps much. The books I most often recommend as Python intros don't seem to touch the subject, and the reference books seem to go not much deeper than the Python docs. Even David Mertz's useful Text Processing in Python doesn't cover i18n (this surprised me).

My recent foray into i18n actually straddled Python and XML worlds. For XMLers, there are a few handy resources:

XLIFF is of particular interest, but I decided not to use it for i18n in 4Suite/XSLT because I wanted to base my work on the Python facilities, which are well tested and based on the de facto standard gettext.

Anyway, am I missing something? Are there all sorts of great resources out there that would slide Python developers right into the i18n groove?

[Uche Ogbuji]

via Copia