"Magnus Manske" Magnus.Manske@epost.de writes:
Well, of course, you can write [[de:talk:Wikinger]] to link to the talk page of the Wikinger article on the German wikipedia!
I think namespace and language should be orthogonal. Unless we can unambigously discern language codes and namespace names, a different delmiter should be used. One could mandate namespaces to have at least four letters, though.
With "automatic collection" I mean this: The U.S. article (should the interneation wikipedia be "us:" namespace??) links to the German article.
Eh, eh, its an /English/ article, so its code should be en.
I think this automatic collection that runs off-line once a day/week is flawed. Imagine pages [[en:A]] and [[de:A]]; someone points to the second in the first. The collection will add a link back from [[de:A]] to [[en:A]]. Later [[en:A]] is split into [[en:A1]] and [[en:A2]] and the link to [[de:A]] is put into A1, not A2. What will the robot do with this situation? What if [[de:A]] were also split before the collection takes place forming something like:
en:A -> en:A1 =====> de:A -> de:A1 =====> en:A -> en:A2 -> de:A2
IMO it is easier to just keep the equivalence information in an area separate from any specific language and always up-to-the-minute. So if someone changes the information "en:A is equivalent to de:A" to "en:A1 is equivalent to de:A" all consequences are propagated immediately.