Re: [lisp] WG Review: Locator/ID Separation Protocol (lisp)

[Robin Whittle <rw at firstpr.com.au>]

I propose changed and additional text for the LISP WG charter in
response to two types of concern:

  1 - LISP's name and claim to implementing true Loc/Id separation
      is problematic, in that other systems such as HIP implement
      truly separate Locator and Identifier addressing systems (AKA
      namespaces) while LISP does not.  [1] [2]

  2 - That the charter should be unambiguous for readers who know
      nothing about the RRG, LISP etc. and should provide greater
      clarity about the nature of LISP's "separation".  This is
      especially important considering the false claims made in some
      LISP I-Ds, presentations and mailing list messages [3] that
      LISP involves separate namespaces for Locators and Identifiers.

   - Robin


Replace the first 2 sentences of paragraph 3 with the following text
(perhaps as one or two paragraphs, rather than the three below) and
leave the remaining sentences as a separate paragraph.  This new text
also explicitly states that EID space is for end-user networks and
lists the benefits it provides in a scalable manner.

   LISP supports the separation of the Internet address space (IPv4 or IPv6)
   into multiple regions, each of which contains only EIDs or RLOCs. EID
   addresses are intended to be attractive for end-user networks and to
   enable the provision of multihoming, portability of the EID addresses
   between ISPs and inbound traffic engineering to large numbers of these
   networks in a scalable manner.

   LISP Ingress Tunnel Routers (ITRs) handle packets addressed to EID regions
   of the address space via global "map-and-encap" system (RFC 1955) in which
   a mapping system is used to determine the RLOC address of an Egress Tunnel
   Router (ETR) to which the packet will be tunneled. The ETR decapsulates the
   packet and forwards it to the destination network.

   Hosts and conventional routers make no distinction between addresses in the
   EID or RLOC regions of the address space.  No separate namespaces are
   created.  While it may be possible to use an IPv4 RLOC to tunnel a packet
   addressed to an IPv6 EID address or vice-versa, IPv4(6) RLOCs and EIDs share
   the same IPv4(6) namespace.

Optionally, at this point, add a mention of HIP as involving separate
namespaces for Locator and Identifier.



I also propose modifying the last sentence of para 1 to include
mention of the proper term which covers LISP, APT, Ivip, TRRP and
Six/One Router: "core-edge separation schemes" [4]:

  In general, these proposals are core-edge separation schemes and were
  initially known by the term "Locator/Identifier separation".



[1] Margaret Wasserman  2009-03-12
    http://www.ietf.org/mail-archive/web/lisp/current/msg00265.html

[2] Fred Templin  2009-03-13
    http://www.ietf.org/mail-archive/web/lisp/current/msg00266.html

[3] Robin Whittle  2009-03-18
    http://www.ietf.org/mail-archive/web/lisp/current/msg00273.html

[4] Towards a Future Internet Architecture: Arguments for
    Separating Edges from Transit Core
    Dan Jen, Lixia Zhang, Lan Wang, Beichuan Zhang
    http://conferences.sigcomm.org/hotnets/2008/papers/18.pdf