On Jun 24, 2007, at 5:21 PM, Matthias Wimmer wrote:
For JIDs: This all has been already solved with stringprep which XMPP uses (i.e. Codepoints after stringprep normalization). For the case of roster items: Characters as how they are sent be user user that creates the roster item (i.e. Codepoints as transmitted).
We don't do stringprep for roster groups. Perhaps we should, but that's a different topic.
I don't see where you have a problem with it. - And if you would have it, you must also have a problem in couting octets, as they very depending on the type of normalization as well.
The reason I have an issue is because there exists logic, database structure, test cases, and the like to deal with JIDs as octets, all of which code can be reused if we pick the same approach. If we pick a different one, even if we think it's better, it will lead to increased cost, complexity, and brittleness in server implementations.
