Joe,
In that sense, I'd view the Internet as supporting overlay networks, in the generic sense that an overlay network running (say) on end hosts could have packets traversing an overlay link be delivered via the Internet.I'm asking whether you then consider TCP and UDP overlays on IP.
No, first, I consider TCP and UDP as protocols, rather than overlays. Also, I don't consider them protocols running on an overlay network because they are aware that they are running on IP and do not themselves provide (say) a network layer. That is, in layering, the focus is on protocols and their relationship with the layer above and below, whereas an overlay (and a virtual network) provide more transparency and can themselves run their own suite of protocols. That said, a TCP socket or UDP socket might be used as a building block for an overlay link in some virtual network (just like a GRE tunnel could be), but that's different than saying TCP or UDP themselves are overlays.
Ethernet supports Appletalk, Decnet, etc. - does that mean that ethernet supports network layer virtualization? that (as above) IP supports transport layer virtualization?
Nope, not to me, because that's really about layering. Appletalk and Decnet here are running on top of Ethernet.
IMO, it's useful to distinguish virtualization from layering; the former ought to add a layer of indirection.
Yup, I agree.
That's what tunnels provide; perhaps that focuses too much on the mechanism
Yup, but I agree it focuses too much on mechanism, and on the virtual links rather than including the nodes.
(I say "network" here because VLANs don't support network indirection; they support only link indirection; endpoints are not abstracted)
Yes, exactly. -- Jen
Note Well: Messages sent to this mailing list are the opinions of the senders and do not imply endorsement by the IETF.