[Standards] Re: IDNA text for rfc3920bis

[Joe Hildebrand]

On Apr 15, 2007, at 12:28 AM, Mridul wrote:

> So in this case, how is S2S expected to interact with this ?
> We will end up with node at sub & node at sub.domain aliasing to the same  
> JID but not so according for remote server(s) processing it -  
> assuming they are able to resolve both to the hosting server  
> (simple matter of dns especially for internal deployments, it will  
> become an A or AAAA record lookup at times).
> Or do we just ignore the problem since it is not the right thing to  
> do ? And expect things to as though both are different jid's to  
> external world and same internally ? kind of breaks a bunch of things.

Agree with Bruce here.  I use a setup with local names often enough  
that I'd like it to be legal.  You just have to be careful what you  
try to interop with, although that leads to Bruce's problem of too  
many unqualified names going to the root for resolution.  It seems  
like the place to solve that is the local DNS server, though.

> Which actually brings me to another question - do we have a concept  
> of JID aliasing in xmpp ? (I did not find any).
> If not, is there any proposal to add it ? This is actually  
> something customers do ask us about - especially since it is quite  
> common in email.
> For example, I would prefer mridul at sun.com while communicating to  
> external user's to <my_unix_id>@sun.com which is my actual JID -  
> but both should resolve to the same canonical JID : though I would  
> like to use both depending on my needs ...
> Especially in context of S2S, there are interesting side effects of  
> trying to support this.

We've shied away from this several times.  The S2S case makes this  
particularly difficult.  The from address has to be something on the  
forwarding server, not the original from address.  This is one of the  
reasons I keep advocating for the addition of "original from" and  
"original to" address types to be added to XEP 33.  Then: what  
happens to probes?  How do you get presence to the other side?  Do  
you need the equivalent of source routing (running everything back  
through the forwarding server) to preserve the illusion for the  
original sender?

All in all, it's easier to write and use an XMPP-XMPP transport.