I am not sure how I never noticed this before, but there is a discussion of SRV records in H.323 that seems to be based on a misunderstanding of the SRV RFC. Also, there was a mistake in the original Experimental RFC that is being fixed in the new internet draft that is just about to become a proposed standard. In brief, the thing that you should query for would be something like _ITUh323ras._udp.x.foo.com or _ITUq931._tcp.x.foo.com (or _udp for Annex E.) The idea is that there are many protocols that can have SRV records at the same place (eg. x.foo.com) in the DNS namespace, so they are further qualified by a protocol hierarchy.

The "_ITUh323ras" part will need to be decided and documented, so this does appear to need to go in the spec somewhere. Understand that I am not proposing that _ITUh323ras be the protocol label used for this, just using it as an example of what is needed.

I am not a member of ITU, but would be happy to work with others who are to get this piece fixed. I am not sure that this can be/needs to be fixed for V3. For those interested in reading the current draft going towards PS, see draft-ietf-dnsind-rfc2052bis-02.txt.

SRV aware nameservers and resolvers are starting to appear now. In particular the latest release of BIND now has SRV support and the Microsoft windows2k nameserver has SRV support and tools using it. ITU support of SRV record usage will help push the draft along the IETF standardization track.

jerry