As requested, I've set up a Dutch test wiki running the new software. I've been sent also a Danish translations file, but I don't yet have access to the usemod database for the current Danish wiki to convert from; when I have that I'll set it up too.
A couple days ago I wrote:
(This will be the same location the final upgraded wikis will be at, but the current .com URLs will also be redirected there once things are set up for real.)
A number of things are probably broken, including:
- There may be searching problems with non-ASCII chars
Please check!
- Article titles which start with non-ASCII chars may not have proper
case conversion. This is either a bug in the locale setup or in PHP's ucfirst() function; if I can't iron it out, I'll copy in the workaround I'd used in an earlier version of the program.
Now fixed on pl, eo. Nl is latin-1, shouldn't have a problem.
- There are some article titles in the Polish database which started
with lowercase non-ASCII chars. These may have been skipped in conversion; this should be fixable with automatic renaming, remind me to do that.
I'm told this is not a problem for the Polish wiki. Please check on others.
- The automatic conversion of uploaded files is very buggy so I
temporarily disabled it; I'll rework that in a bit. In the meantime, test to be sure that uploads work.
More or less fixed now. Please check them!
Files previously linked from the English wiki are automatically copied over in the conversion and relinked with the new syntax; files linked from the meta wiki are not, mainly because that's on a different server and I don't have clean access to the upload space there.
- User accounts are not carried over; make sure you can create and use
new accounts.
Still true.
- X-system conversion for the Esperanto wiki isn't currently
implemented; until I get that cleanly worked in, use Ek! (Windows) or an appropriate multilingual keyboard layout (Mac, Unix) to type supersigns.
Still on the agenda...
- I've converted both of these wikis to standard UTF-8 (from modified
UTF-8-with-X-surrogates for eo and ISO-8859-2 for pl). Difficulties internal to the system (for instance, capitalization issues) should be sorted out shortly. Polish folks; if this is a real problem with browser support (can you read & write as before?), let me know; in the worst case I can shift you back to ISO-8859-2 and have more charsets to deal with.
No one's complained to me yet (though there are some URL-prettification issues in general which I will be attacking when I get a chance; especially as regards the note at the bottom of the Printable version), so I guess that's fine. Dutch and Danish will continue to be ISO-8859-1 for the time being.
- URLs with titles in the old syntax/charset aren't compatible yet,
this will be added in before final conversion, so URLs will be preserved. On the same note, interlanguage links to English, French, German etc (ISO-8859-1) with accented characters won't work if typed in straight. That'll get fixed in a bit.
Still on the agenda. Dutch, Danish are the same as English, German; don't have to worry about this.
- Namespaces may be generally broken, and some translations may suck.
:) Suggestions for improvements and corrections are welcome.
Please check these!
- Talk (diskusja, priparolu) links are (or at least should be)
automatically converted to namespaces, but the links aren't taken out of the body text yet. That'll be done later.
These should now be automatically plucked out in most cases (where the link appears alone on a line and it looks safe to remove). Please check that real links haven't been disappeared too.
Note that pre-conversion text should be preserved in the edit histories, so individual problem conversions should be easy to restore. Serious problems, of course, should be fixed in the script before the final conversions are made.
- The timezone on the new server is UTC instead of California time;
for Polish eventually the _default_ offset will be local Polish time, but this isn't ready yet.
Done. Polish, Dutch should be on Central European time by default, as is the live German wiki. Manual timezone offset is still against UTC.
I'll be making improvements over the next few days; if things are looking reasonably good, expect final conversion by the end of the week. Please play with editing, special pages, etc and complain at me if things other than the above-mentioned don't work.
Oh, and here's the page on the English wiki which describes the various changes from the old software to the new: http://www.wikipedia.org/wiki/Wikipedia:PHP_script_new_features
-- brion vibber (brion @ pobox.com)