Two weeks agos, fr decided to go to UTF-8. Now we can say, that fr is working fine :)
Some problems we had :
- Pbs with illegal windows caracters (which are not in ISO8859-1). Corrected by a bot
after the conversion, but It could have been done before.
- Cache pb, anonym users created pages with old encoding. If we do another conversion,
we'll have to clear the squid cache I think.
- Name of pictures. The name of pictures are converted in UTF-8, but not the file name on
server. Now it's fine, we have a script which convert it.
It tooks around 3 hours to convert the french wiki. During this 3 hours, 20' to make
the dump, 2 hours to convert the dump, and 40' to get back the dump in the database.
The fr site was offline only for the 40' of getting back the dump in database. During
the rest of the time, the wiki was in read only.
I would like to thank Med, for his conversion software, Looxix for his small python script
to convert the pictures file names and all the work he had done to convert illegal windows
caracters with his bot.
If some people want to go to UTF-8, I can say we're ready now :)
Shaihulud