Erik, EN> I don't think an identifier is necessary in every packet. EN> But I do think it makes sense to have a shim layer above IP which EN> uses locators in the packets below (for IP's routing) and presents EN> fixed length identifiers in the pseudo-headers passed to/from the upper layer EN> protocols.
I believe you have described the common characteristic to the set of solutions I classed as "IP Endpoint" in the -analysis- paper I issued last night. It distinguishes a "shim" approach from an "ip-es" approach with a shim being more backward compatible for the upper layers, but that is a nicety, rather than a key point, I believe. EN> The reason having identifiers in every packet isn't useful is that if EN> you want to avoid facilitating redirection attacks of packet flows, EN> then the receiver needs to verify at some level the relationship between EN> locators and identifiers. yes! EN> Thus between the ULPs the shim provides a service which EN> passes what looks like packets containing 128 bit identifiers, even though on EN> the wire the packets have 128 bit locators. Yes. I've come to the view that that is really what MAST (and I believe LIN6 and HIP) are trying to do, though I originally described it in NAT terms. 128bits vs. 32 is a v4/v6 distinction, not an architectural distinction. The point is that transport sees something that is like what it is used to, but it no longer is really a locator. Instead it is an identifer that the new layer (shim, ip-es, or whatever) maps to the locator. d/ -- Dave Crocker <dcrocker-at-brandenburg-dot-com> Brandenburg InternetWorking <www.brandenburg.com> Sunnyvale, CA USA <tel:+1.408.246.8253> -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------