On Tuesday, 29th August 2006 at 20:58:48 (GMT +0100), Rob Church wrote:
it could well be that MediaWiki is expecting to find latin1 data, in which it wraps UTF-8 data regardless, and does some encoding/ unencoding work each end. If it doesn't know that the database table it's reading is using UTF-8 collation, then it is bound to bugger up the usernames, no?
Well... the phpBB database is, in fact, *not* using UTF-8 collation... Looking at the tables in phpMyAdmin, the collation still says "latin1_swedish_ci" as it did before. However: to turn phpBB into a truly multilingual board, we had to (among other things) download an SQL dump of the entire phpBB database, then manually convert that file into UTF-8, then upload all the tables back into the database.
This was the final step necessary to make phpBB truly multi-lingual. If left out, gibberish was displayed. However, this final step now seems to confuse MediaWiki when it tries to fetch users' names from the table. (Our effort to bridge phpBB with Coppermine failed as well, perhaps for the same reason... However, we prefer a fully multilingual albeit isolated board over one that would be integrated at the cost of being mutilated.)
Maybe the issue will be resolved on its own when phpBB rolls out one day with UTF-8 as its default encoding and manual conversions are no longer necessary.