On May 13, 2011, at 13:02 , Glenn Maynard wrote:

> On Fri, May 13, 2011 at 1:54 PM, Matthew A. Miller 
> <[email protected]> wrote:
> Yes, "service domain" == "domainpart" in this context.  This the general 
> consensus for all XMPP authentication.
> 
> Is there anywhere these general consensuses are documented?  As an 
> implementor, it's painful having to wade through mailing lists trying to 
> figure out these things that aren't fully specced, and as a user it's hard to 
> report bugs when there's nowhere authorative (much less normative) to point 
> to in bug reports.
> 

This used to be sort of documented in a "Tao of XMPP" wiki site, that I can't 
find again. /sigh

> Looking briefly at what Pidgin's XMPP implementation passes to Cyrus, 1: for 
> direct, non-SRV-discovered XMPP, the JID domainpart and the hostname are the 
> same it uses the JID domainpart (which is the same as the hostname in this 
> case anyway); 2: for direct, non-SRV XMPP where the user has specified a 
> different XMPP to connect to than his JID domainpart (eg. Gtalk), it uses the 
> specified hostname; 3: for SRV discovery, it uses the SRV-discovered 
> hostname; 4: for BOSH, it uses the hostname of the BOSH server.  So, not all 
> implementors seem aware of this consensus...

This is the second biggest reason why SCRAM is better than DIGEST-MD5.  I will 
note that many server-side implementations of DIGEST-MD5 don't try to validate 
the service URL.


- m&m

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to