[xmppwg] S2S and SASL (xmpp-core 6.1/11.3)

[Eric Rescorla]

Marshall Rose <mrose+internet.ietf.xmpp at dbc.mtview.ca.us> writes:

> > Joe Hildebrand <JHildebrand at jabber.com> writes:
> > 
> > > We decided the last time we beat this horse to death that dialback adds nice
> > > deployment choices for non-trivially-sized server installations, so that the
> > > inbound side might be handled on a different set of servers from the
> > > outbound side.  Without dialback, we lose this flexibility, which *has* been
> > > useful in real-world deployment.
> > > 
> > > That's why I *thought* we had already decided to leave it in.
> > Hmm... I've never heard this rationale before.
> > 
> > In any case, this has nothing to do with any "authentication"
> > purpose, so it probably shouldn't be called authentication.
> 
> well, i heard that and about 50 other things back in san francisco.
> 
> why don't we just move the thing into an appendix, change "authentication" to
> something which isn't someone's hot-button, and move on?

The thing needs a reason for living.

Either it's:
(1) A security mechanism, in which case it needs an explanation of
    how.
(2) A way to have split in/out, in which case the document needs
    to say it.
(3) Some deprecated way of talking to legacy stuff, in which case
    the document needs to say that.

Protocol features shouldn't just be floating around with no particular
justification.

-Ekr


-- 
[Eric Rescorla                                   ekr at rtfm.com]
                http://www.rtfm.com/