Edward Z. Yang wrote:
Hi, I'm currently developing an I18N system for a
moderately large PHP
software library and would like to model it off of MediaWiki's setup,
being an extremely well-tested installation (even though it is an
application, not a library).
Will this library be free?
However, I do realize that the stability that comes
with age also means
that there might be a little bit of legacy cruft stuck in the files,
esp. when concerning fundamental architectural changes.
I'd like to ask all and any developers who are familiar with MediaWiki's
I18N structure to comment on your opinions of the strengths and
weaknesses of the system (the code will tell me how it works) and what
you would do different if you had the chance to rewrite the whole thing.
Consider it documentation for an extremely undocumented section of the
code: not a single result turns up for a Wikitech-l search of "i18n."
I've made a number of posts to wikitech-l on this subject, but I don't use
the term "i18n". As I once quipped on IRC, MediaWiki isn't
internationalised, it's international. It's always been international, we
never had to go through the process of internationalisation to get there. Or
more precisely, it's multilingual. But also I don't like substituting digits
for letters.
There are a number of aspects to our interface translation system:
* Message input
* Architecture, caching
* Parameter substitution: agglutination and escaping
Which of these are you interested in? Presumably our message input model
(i.e. via a wiki) wouldn't be universally applicable.
-- Tim Starling