A question about validating alphabetical data

[eluser]plainas[/eluser]
Hey all, charsets and such stuff is one the most confusing subjects for me.

I need to validate data that will be part of an url.
The users will be able to input a fragment the url of the pages they can create.

for example

a user will be able to create this kind of url:

http://mydomain.com/myapp/username/user_created_data

Now, to be safe, I was thinking of allowing only a-zA-Z0-9\-_ chars, but I will have many users that will generate content in several languages.
My biggest problem is, using esoteric UTF-8 enconding, How do I sort out among hundreds of kinds of chars which ones represent only letters/algarisms?
Like, I don't want to allow ponctuation and such stuff to be used, but how do I sort it out in chinese, corean, bangla, etc. chars?

Another question... how do i safely insert UTF chars in an url? is urlencode() enough?