What we need to do is make a solution for this that works. We could of course just encode the characters like the norwegian characters æøå can be used in an URL by encoding it as %E6%F8%E5, but I think that this looks very dirty and is not very user friendly - which was the Idèa with the url_alias.
The second solution which we can implement is a translation table for every character set which will convert o->oe, æ->ae and å->aa. But this will only work for character based languages.
For chinese and japanese I do not have a good suggestion.
A fallback solution would be to use the node id if the name is empty. E.g. /home/news/120/420/.
I agree with you .. will be to long and ugly link :)
But .. I have some sites which use friendly url, and it`s can grap the UTF8 fonts normal, but the i don`t use utf-8 charset, am using windows-1256, and it`s work fine, and geting the artical from DB by the artical name which is in utf8 , but will be little bit difficult to write the link manually for hebrew and arabic language, beciz is starting from right to left :) !
We are about to set up a Hungarian language site, where I'd like to use a translation table like solution for converting accented vowels:
á -> a
ó,ö,ő -> o
ü,ű -> u í -> i
Since the existing url_translator simply leaves these above mentioned letters out from the string, the practical result is something "very difficult to understand" urls.
I understand, that implementing of this charactertable conversion module is not top priority for eZ, so I am happy to do it (or at least try :), also I am not very familiar with the source yet, so I'd be glad for some assistance of locating the script responsible for URL translations.
Is there any suggestion how editors can assign url_alias manually for each object?
For exapmple: Directory "Economy and business" has url_alias "business" only and an article with title "McDonald's, Wendy's shares beef up" has url_alias "sharesbeefup".
Lazy editors can simply leave this field blank to generate it automatically.
And, finally, I don't think that translation table ф,Ф->F should be good enough for non-latin languages (russian, chinese, bulgarian, etc) because such translation kills the sense constituent of "url_alias"
BTW: Nice urls are important not only for users, but can bring some additional pageranking for search engines.
I agree with eoNormalMutation, as for non-latin based Language, manually alias_url is the best choice (in fact by fat the only solution I can come up with)
I did patch eZ 3.1 to support this, although with some limitations( only main location's url can be assigned).
Unfortunately, I lost all the source code :(
Hope this would be a new feature in eZ 3.4 series.