On 7 March 2010 10:42, Evgeniy Khramtsov <[email protected]> wrote: > Matthew Wild wrote: >> >> People want this, it's trivial to do, we should >> standardize a way of doing it. Done. >> > > Actually there is already a standard for address discovery - STUN.
STUN != XMPP. > By the > way, you didn't tell why this XEP is useful and why it is better than STUN, > or when to use this XEP and when to use STUN. I don't necessarily want STUN support - it does more than I need. My server may not support it, whatever, I don't care. >> Now this XEP isn't telling people not to use STUN, TURN, UDP or >> Jingle... it's for the people who don't want or need to use those >> technologies (perhaps for the moment). > > This XEP adds incomprehension: a Jabber developer might be confused which > approach to choose: this XEP or STUN. I highly doubt that - STUN is referenced explicitly in the XEPs that require it. >> >> I don't feel we should be >> limiting what people want to do with XMPP, or how they should build >> their applications > > Agreed. But I think we need to use a common well-tested solution, and not > invent our own without necessity, if possible. > A simple request/response protocol is hardly "inventing our own". > BTW, STUN has a protection from poorly written ALGs which rewrite IP > addresses, that's why MAPPED-ADDRESS is now deprecated in favor of > XOR-MAPPED-ADDRESS. This XEP doesn't have such protection. > Indeed, and I argue that it shouldn't. Matthew
