A comment in JEP-0156 confuses me: 3.1: A domain SHOULD NOT present information in DNS TXT records that is available via the DNS SRV records defined in RFC 3920.
At first I read that as '_xmppconnect IN TXT "_xmpp-client-tcp' shouldn't be allowed, because the SRV records advertise the same data, or am I mis-reading something here? Can you give an example of data in SRV that should not be included in TXT? Is the _only_ case port numbers? Surely it would actually be a _good_ idea to include the port numbers in the TXT record, bcause you reduce the number of DNS lookups? (i.e. if you get a TXT record, don't bother with SRV) On 11/17/05, Peter Saint-Andre <[EMAIL PROTECTED]> wrote: > Ian Paterson wrote: > >> While handling the route attribute, should the authority > >> component of the IRI be used or ignored? > >> > >> What's the suggested result when the IRI holds no node identifier? > >> Should the route attribute be silently ignored, or should an error > >> (improper-addressing seems suitable) be thrown? Is it safe to try to > > use > >> the authority component address as an last-resort solution in such a > > case? > > > > The JEP states that the XMPP IRI indicates the "protocol, host, and > > port". Although the current version of the JEP does not currently > > explicitly exclude other IRI components, perhaps it should. The XMPP IRI > > SHOULD be of the form: > > "xmpp:" ihost [ ":" port ] > > > > Can anyone think of a use case that would be prevented if we formalise > > this in the JEP? If not then I would say that 'route' attribute values > > with a different form SHOULD be silently ignored. > > I see no need for including anything but xmpp:ihost[:port] because the > whole point here is specifying which server the proxy will talk with. > > Now that we have JEP-0156, do we need the :port in IRIs for this use > case? That is, can't the proxy figure out which port to use via DNS TXT > records? Does the client really need to tell the proxy which port to use > or is that task better left up to the proxy? Just asking. > > > Also the JEP states that "The XMPP IRI specifcation does not currently > > allow a port in an XMPP IRI; the authors will pursue the matter within > > the Internet Standards Process." I'd like to fix both these points at > > the same time. Peter, is there any news about the possibility of > > including ports in an upcoming draft-saintandre-xmpp-iri-03.txt? (IIRC > > this was discussed on the Standards-JIG list a few months ago.) > > I wonder if we really need to specify the destination server as an IRI. > What do we gain from using URI/IRI syntax? Why not just specify > host:port since the protocol (xmpp) will always be the same? This is for > use strictly within the context of XMPP so the usual arguments about the > need for a URI/IRI don't apply (identifying XMPP entities from outside > of XMPP networks). It seems simpler to just specify host:port in the > 'route' attribute and be done with it. Also, that way I don't need to > add :port to the XMPP URI draft, which still concerns me a bit because > ports are not part of the base XMPP address spec. > > Peter > > -- > Peter Saint-Andre > Jabber Software Foundation > http://www.jabber.org/people/stpeter.shtml > > > -- - Norman Rasmussen - Email: [EMAIL PROTECTED] - Home page: http://norman.rasmussen.co.za/
