Some cautionary comments

[Oliver Spatscheck <spatsch@research.att.com>]

Mark Day writes:

 > 1. Negotiation of parameters
 > 
 > I think the most elaborate supportable model here is where *one* side of a
 > communication presents *one* set of choices to the other side, which then
 > picks *one* from that set and the negotiation is over for the life of that
 > communication between those parties.
 > 
 > I'm willing to be convinced otherwise, but I want to avoid casually assuming
 > that negotiations are a good idea or easily built.
 > 

Mark,

I think this is a question of how tightly the advertisement is coupled with
the selection. If as you describe the selection follows the advertisement
directly I agree with your two phase approach. However, my understanding so
far was that the advertisement is sent to many receivers which can potentially
select. In this case I strongly believe that a third phase is needed in which
the advertiser acknowledges the selection.

At this point I don't think I am sure if we want to advertising a set
of choices to one or many receivers at the same time. However, I wanted
to point out  that if we decide to advertise it to many receivers at the
same time (like BGP) two phases won't do.

Do you envision us advertising a set of choices to many parties at the
same time or do you think that would be to complicated?

Oliver