On 12 March 2010 11:57, Dave Cridland <d...@cridland.net> wrote: > On Sat Mar 6 09:33:25 2010, Pedro Melo wrote: >> >> Besides, this is a trivial XEP. The C2S already has your IP address, >> so its easier to ask your server for it. > > That's not actually clear to me. > > In the majority of our deployments, M-Link sits in a DMZ, with routable > access to client IP addresses, such that it sees internal addresses rather > than external. I have no reason to think that in this respect, our customers > are different from anyone else's. Since the majority of the clients connect > from internal addresses, it follows that this protocol will return the wrong > address in the majority of cases for the majority of deployments. >
I'm not sure I'll agree that it'll affect the majority of users without seeing figures (that I know full well can't be produced) :) > I understand the need for this XEP, but it would be useful to acknowledge > that STUN exists and is preferable (since STUN servers will be explicitly > located to avoid the aforementioned problem) - the current XEP doesn't have > any mention of STUN, which is a little alarming. > +1. > For one thing, suggesting that this will usefully indicate to clients that > they are located behind a NAT isn't true if the server is also located > behind the same NAT... > +1. I think this needs clarification. So let's get the people who have been pushing for this XEP to describe their use-cases. Matthew
