Mailing List Archive

On dim, 2002-03-10 at 06:34, Tomasz Wegrzanowski wrote:
> Everyone on these lists who can speak Polish - please check this translation
> and whether internationalization works in general. Not all strings have been translated yet.

Great! I don't speak but a few words of Polish myself, but I'll take a
general look over it this evening.

> It uses UTF-8.
> I think that's the right way.
> If you think otherwise please say why.

I agree.

> 1) Print footer ("This article is from Wikipedia (URL), the free online
> encyclopedia. You can find this article at URL") is in wikiPage.php
> and it should be in wikiText*.php
> That is it should be internationalizable.

Huh. Well, that'll need to be fixed...

> 2) The same word is used for names of languages and names of non-English
> Wikipedias. That's wrong for Polish because "language" is masculine
> but "Wikipedia" is feminine, so the word should be different
> ("polski" and "polska").

Where exactly does this come up? The only place in the interface that
the languages are displayed (that I can think of) is in the link bar
with links to other languages (ex http://wikipedia.com/wiki/Esperanto).
As it is, these are going to need to be shortened, probably to
2/3-letter language codes, because the line gets waaay too long with
many languages.

> 3) There are problems with interwiki links. If Polish article title contain
> non-ascii characters, I can't easily link it using [[pl:Name]].
> The "right" solution is of course using UTF-8 for all wikipedias.
> We will have to do it some day, as not doing it is causing
> more problems than it saves. But that's not an urgent issue.

Yes, that's what we're planning.

-- brion vibber (brion @ pobox.com)

On Sun, Mar 10, 2002 at 08:40:34AM -0800, Brion L. VIBBER wrote:
> > 2) The same word is used for names of languages and names of non-English
> > Wikipedias. That's wrong for Polish because "language" is masculine
> > but "Wikipedia" is feminine, so the word should be different
> > ("polski" and "polska").
>
> Where exactly does this come up? The only place in the interface that
> the languages are displayed (that I can think of) is in the link bar
> with links to other languages (ex http://wikipedia.com/wiki/Esperanto).
> As it is, these are going to need to be shortened, probably to
> 2/3-letter language codes, because the line gets waaay too long with
> many languages.

My mistake. $wikiLanguageNames is used only for links to other
language versions (with [[pl:Foo]]). So it should just use one form.

The second place where it appears - links on main pages - are not automatic.
So this isn't any problem after all.

On dim, 2002-03-10 at 09:54, Jimmy Wales wrote:
> pl.wikipedia.com
>
> How about I set this up on Monday? Do I need a new conversion script?

I'll need to tweak the conversion script to convert the Latin-2 database
to UTF-8, and drop in a quick fix to make searching work on non-ASCII
words. Shouldn't be hard, I'll drop a note when it's checked in.

The full set of encoding-related changes to the code still need to be
determined and made, but if we start it off using UTF-8, the database
should be ready.

(Also, we may need to rebuild the search index later, as that part of
the code may still change. But then this is the case if we change the
mysql indexing anyway.)

> I believe that this is best:
>
> 1. I set up pl2.wikipedia.com as the "beta test" Polish wikipedia,
> and import the existing data from pl.wikipedia.com using a conversion
> script.
>
> 2. We run this for one week, inviting as many Polish speakers as we
> can find to test it, incorporating bug fixes all week as quickly as
> they come in.
>
> 3. The following Monday, I make the conversion "for real".

Sounds good to me. Can we commit to such a schedule for Esperanto as
well? We've only put it off so far because we were going to change the
database internals, but since we seem to have decided on UTF-8 it's not
a problem to convert ASAP and make the changes later.

> I'd like to follow up with Spanish and Esperanto as quickly as possible.

FYI, I currently have a beta running the converted Esperanto wiki at
http://leukas.dyndns.org/wiki/ . A more officialish server wouldn't hurt
though, as this is limited to the reliability of my DSL line.

-- brion vibber (brion @ pobox.com)

On dim, 2002-03-10 at 08:40, Brion L. VIBBER wrote:
> On dim, 2002-03-10 at 06:34, Tomasz Wegrzanowski wrote:
> > Everyone on these lists who can speak Polish - please check this translation
> > and whether internationalization works in general. Not all strings have been translated yet.
>
> Great! I don't speak but a few words of Polish myself, but I'll take a
> general look over it this evening.

Okay, it looks pretty good to me! I can't speak for the correctness of
the translations, of course, but most obvious functions seem to work and
I didn't see gratuitous English text in the main places. I've checked it
into the CVS server. (Yay!) I'm not sure if all the log names are right
(eg, Log:Uploady), some of those may be hardcoded English internally.
(But, "log" shouldn't be relied on anyway.)

Also, test that all the talk/dyskusja pages work correctly; currently
there's some magic in places that expects all talk pages (including user
talk, wikipedia talk, log talk, etc) to have the same word as just plain
talk pages in them. Right now you don't have that, and it doesn't seem
to like it. ie, we have a user page in the "Użytkownik:" namespace,
which goes to "Użytkownik Talk:" (which it won't let me create) rather
than something like "Użytkownika Dyskusja:" (let's pretend that was a
correct genitive case or something). However, this whole part of things
is rather rough now, and needs work. (My mind is going, can you tell?
It's too late to be working on software...)

I also added a quick Latin2-to-UTF-8 convertor to convertWiki2SQL.php;
uncomment the Polish section at the top and it'll enable it (as well as
the alternate field separator and "Dyskusja" as the default /Talk link).

A couple of warnings; the conversion script currently doesn't fill in
link information, since the entries in the cur table were removed.
Support should be added there for filling the new linked and unlinked
tables, but for now you can just run rebuildLinks.php after creating the
new database.

Also, searching with non-ASCII characters is still broken until the much
discussed, not yet implemented second index field is added. (Or we
figure out how to make MySQL behave without one.)

-- brion vibber (brion @ pobox.com)

On dim, 2002-03-10 at 09:53, Tomasz Wegrzanowski wrote:
> My mistake. $wikiLanguageNames is used only for links to other
> language versions (with [[pl:Foo]]). So it should just use one form.
>
> The second place where it appears - links on main pages - are not automatic.
> So this isn't any problem after all.

Ah, good.

I have fixed the print footer -- set the non-English message in
$wikiPrintFooter.

Also, the conversion script now builds the link tables, so running
rebuildLinks.php is no longer necessary. Next step: bringing in the old
edit history!

-- brion vibber (brion @ pobox.com)