On 2010-09-14, Mikael Abrahamsson wrote:

I don't believe in overlay networks. It's generally expensive to do encapsulation/decapsulation, I've always avoided it as much as been possible.

It should be avoided as much as possible, but then, it usually cannot be fully avoided thanks to layer-to-layer translations. Take for instance DNS; it's a full-blown overlay in itself already, yet a necessary one. It's just that we shouldn't be building an overlay on top of another overlay without a very good reason.

I think the way towards such a goal is to build more expressive interfaces between the different layers, as need arises. That way functionality doesn't have to be rebuilt at higher and/or inappropriate layers, since everything that is needed to implement it at the proper layer is already present.

Of course such interface/layer definitions are hard to do. But not *that* hard at the edge. They tend not to involve hardware today.
--
Sampo Syreeni, aka decoy - [email protected], http://decoy.iki.fi/front
+358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to