> From: Keith Moore <moore(_at_)cs(_dot_)utk(_dot_)edu>
> when that functionality requires having knowledge that is only
> possessed by the network (which is what hosts need to do address
> selection)
Rome wasn't built in a day. If you have a solution that requires pieces
A, B and C, saying you can't deploy one unless the others are already
deployed is a recipe for statis. Yes, "the network" does need to make
more info available to the edges, but that's another issue.
> (you might make an argument for moving the functionality to the
> edge of the network, but not all the way to hosts, and that
> argument might be convincing)
In general, in these discussion I am not differentiating between "edge of
the network" and "hosts", because the important thing is that it's not
*in* "the network" - i.e. completely unavoidable by all users. Congestion
control was in "the network" in the ARPANet, and routing is still in "the
network" today.
Once you have moved things "out of the network", it's a second order
decision whether you do it in the hosts, or whether you have some
intermediate set of substrate (e.g. DNS) at the "edges" (from the point
of view of "the network") that does it.
(Note that the substrate is something which you may well also wind up
depending on for the proper functioning of the network - e.g. because,
let's be serious, if DNS dies, for most users, the network is dead.)
Noel
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf