--On June 23, 2010 10:07:44 PM +0200 Patrik Fältström <paf(_at_)cisco(_dot_)com>
I did chat with a few server implementors about this and the feeling was
SRV + .well-known is a good solution that can quickly be deployed. Some
1) SRV's are very deployable today - a new RR will be harder to deploy.
2) There is a push to support SRV's with other protocols (e.g.
draft-daboo-srv-email for IMAP, POP3, and Submission) so from an
operational standpoint its likely to be more common.
3) .well-known is useful in the absence of any DNS records. i.e. if no
SRV/URI were available, a client can still try auto-discovery by
attempting an HTTP connection to the host (derived from user input) and
the .well-known path.
Hmm...regarding the new RR, the only thing I can think of today is the
need for some changes in the provisioning system from which one create
the DNS zones. I do not know of any DNS code today that can not handle
"unknown" DNS RR Types, but maybe I am wrong? I am though confused over
(1) as this is 2010 and not 1998.
I am thinking more of client APIs.
Regarding (3), I think it is sad people move redirects and lookups and
functionality to be on top of HTTP instead of IP. And I do not understand
"in the absence of DNS"...that is an interesting situation ;-)
Specifically to use that as the foundation for the architecture to use
for calendaring, that is -- I hope -- one of the more fundamental
applications on the Internet.
Sorry - bad wording on my part. What I meant was "in the absence of a
service-type to hostname mapping in the DNS".
Ietf mailing list