There's two cases I can think of for this: 1) A restricted client may not have suitable unicode tables to hand, etc.
I think this is discussed in the protoXEP. 2) A restricted user may not have the ability to type the correct jid. This doesn't seem to be addressed, but could be with minor modifications. For example, if I have the ascii-fied jid of [email protected], I could have my client request the correct spelling, in effect, and get back remko.tronç[email protected]. Presumably in this case, it would need to request this from swift.example, and not the client's local server - though it's possible that the local server might forward along the request. This becomes rather more useful for asian jids, where the user may well have both a ascii (and possibly even European) variant of their name as well as the correct script. I vaguely recall Kurt had some ideas along these lines, at summits past. In any case, I'd feel happier supporting a protocol that handled both use-cases, though supporting just the first seems to do no harm. As one comment, having the server offer clients the service on their own jid seems cleaner here, as it'd allow them to set no addressing on the stanza. Dave.
