I think our general consensus is that there should be an article titled without diacritics, or otherwise Anglicized ("Goedel", for example), and that this article should be a redirect to an article with the real title, diacritics and all.
This would work in a perfect world in which the person making the diacriticaly named article also made an ASCII redirect it, Lee. But unfortunately this often isn't the case. Even worse is the fact that most netcitizens don't know how to use their keyboard to make diacritics -- meaning most people searching Google won't use them and that linking directly to these articles is unnecessarily difficult.
That hasn't been a problem in practice; there seems to be a natural urge among some folks to fix such things when they occur, usually because one of the authors is obsessive about the diacriticals and another is obsessive about Anglicization. If we have been slipping up in our Anglicization task, we should simply pay more attention.
The more serious problem is the lack of standarization in HTTP POST data, requiring the software to make assumptions that might be wrong about the character set used for POST data. That's probably the main reason we should be sure to have Anglicized titles, but it's no reason not to also have native ones.
However, I proposed a similar anti-diacritics query on [[talk: naming conventions]] some time ago but soon began to realize that this is a technical issue that can be solved by the wikiware gurus. All that is needed is for the software to treat the diacritic form and the ASCII equivalent the same. For example; é and e could be treated as the same character for searches and for linking (the Google issue might pose a problem though....). That way when somebody searches for [[José Ramos-Horta]] by entering in Jose Ramos-Horta in the wikipedia search engine they won't come up empty handed if nobody made an ASCII redirect to the diacritic name.
That would actually be quite easy; but it would also be wrong. Some languages don't map meaningfully that way, or Anglicize by other methods, and we don't want to make the search less useful for people who _do_ know how to use their keyboards. The right thing to do is make sure that ALL variations on any name appear in the database, either as titles or in the text.
We actually did hash this out quite a bit, and I'm pretty convinced that the present convention is the right thing. If some users don't pay close attention to it, then help them out. We can always add features to the search function later, but it would be missed opportunity to throw away useful information when we gather it just because some people have a hard time dealing with it.
0