Robin, > -----Original Message----- > From: Robin Whittle [mailto:[email protected]] > Sent: Thursday, March 11, 2010 5:27 PM > To: RRG > Cc: Templin, Fred L > Subject: Re: [rrg] LISP does not implement Locator / Identity Separation > > Short version: Fred agrees that LISP is "not really achieving > a true locator/identifier split.". He also > thinks that "the LISP acronym itself seems a bit > misleading". I think it is very misleading. > > We discuss the meaning of "EID" and whether > such addresses - or any global unicast address - > identifies a host or just an interface on a host. > > > Hi Fred, > > Thanks for your reply, in which you wrote: > > > As I have said many times here on the list, what LISP > > is calling "EID" is nothing more and nothing less than > > an IP address. > > Yes - all hosts and all routers except ITRs treat them the same as > any other global unicast IP address. ITRs have a special algorithm > for handling packets whose destination address matches an EID prefix. > > > > IP addresses are assigned to interfaces; hence, they name an end > > system *interface* and not the end system itself. > > > A host may have multiple interfaces, but no two hosts share the same > interface. So if a global unicast address IP address identifies an > interface, it is also true to say that as a result of this, this IP > address also identifies a host. (This is not counting anycast, of > course.) > > > Moreover, these LISP EIDs are routable within a certain scope; even > > if that scope is only node-local. > > I am not quite sure what you mean by "node-local" in the context of > routing packets. > > A packet whose destination address is a LISP EID is perfectly > routable anywhere in the world, assuming there are one or more Proxy > Tunnel routers advertising the "coarse" prefix which covers this EID > address. (In Ivip, it is DITRs advertising Mapped Address Blocks - > MABs.) All the routers except ITRs use the existing algorithm they > use for all other packets with global unicast destination addresses. > ITRs use a second algorithm. For a full description, please see: > > Re: [rrg] LEIDs, SPI & ordinary IP addresses as both IDs & Locs > http://www.ietf.org/mail-archive/web/rrg/current/msg06082.html > > > > Conversely, a true identifier (like a HIP HIT) is not routable > > within any scope. > > Yes, a HIP HIT is purely an Identifier. > > > > So, I have to agree that LISP is not really achieving > > a true locator/identifier split. > > OK - thanks for this. > > > > The LISP EID IMHO would be more accurately renamed as "Endpoint > > Interface iDentifier", and the LISP acronym itself seems a bit > > misleading. ["iDentifeir" typo corrected.] > > Since a group of completely different architectures (Core-Edge > Elimination) do, as the central element of their architecture, > implement the "Locator / Identifier Separation" naming model (which > predates "LISP) I regard LISP's current name as being completely > misleading and a source of difficulty for anyone trying to understand > the "LISP" architecture, or the entire scalable routing field. > > To call what is currently known as a "LISP EID" address as: > > Endpoint Interface iDentifier > > wouldn't make much sense to me, since this description also fits any > global unicast address which a host is using, or any other address > such as an RFC 1918 or RFC 4193 ULA address which also identifies a > host interface, and therefore also a host.
I will confess that I was just trying to make the best out of a tenuous terminology when I came up with this expansion for "EID". But, I think we are pretty far down the road with EID/RLOC to consider abandoning them now. Thanks - Fred [email protected] > In Ivip, I chose the term "Scalable PI" (SPI) address for the new > "edge" subset of the global unicast address space which can scalably > provide provider-independent address space to end-user networks for > the purposes of portability, multihoming, inbound TE and mobility. I > don't have a particular term for the remainder, which remains as > "core" space, according to the meaning of "Core-Edge Separation". > > Since Ivip doesn't slice up SPI space into prefixes, there's no such > thing as an SPI prefix - other than a Mapped Address Block, which is > the DFZ-advertised prefix covering a block of SPI space. I use a > term first used by Bill Herrin - "micronet" - to refer to the integer > number of contiguous IPv4 addresses or IPv6 /64s which all have the > same mapping. > > I think the terms LISP, EID and RLOC are confusing. "Proxy Tunnel > Router" is a poor term too, since it doesn't proxy for anything. > > The term "LISP" (Locator Identity Separation Protocol) is really > muddying the waters and I think the field and the "LISP" architecture > would be much better off if it had a new name. > > Perhaps I will refer to it as: > > "the architecture which, ideally, would be formerly known as LISP" > > - Robin _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
