Dear program committee and all,

shortly before the previous RRG meeting in Dublin, I had proposed a new
host stack architecture as one step towards solving the routing
scalability problem.  The idea was to move the existing indirection
between hostnames and IP addresses from its current position at the
application layer down to the IP layer.  Applications and transport
protocols would thus exclusively work with hostnames, and IP addresses
would be dealt with only at the IP layer.  I would like to let the group
and the program committee know that I am now working on a concept paper
describing this new stack architecture in more detail, and that I would
like to present this at the RRG meeting in Minneapolis.  I hope to have
the paper ready within a few days, as I am aware that this is a
prerequisite for RRG presentations.  Let me for the meantime point back
to the email in which I sketched the new stack architecture initially:

  http://www.ops.ietf.org/lists/rrg/2008/msg01912.html

Like other host-based solutions, the proposed new stack architecture
provides a basis for efficient multi-homing, and it simplifies
renumbering by eliminating the need to renumber hosts and applications.
Unlike other host-based solutions, the new stack architecture has
additional advantages in terms of human-friendliness, low stack
complexity, and simplified application programming:

- Human-friendliness, because the role of a host identifier is taken by
  human-readable and memorizable hostnames rather than by cryptographic
  or encoded bit strings.

- Low stack complexity, because no extra identifier/address split is
  needed, and because IP-address-specific functionality (hostname
  resolution, IP destination address selection, NAT traversal) is
  centralized at the IP layer, avoiding the need to replicate this
  functionality in applications.

- Simplified application programming, because the existing,
  IP-address-oriented socket API is replaced by a simplified API
  providing "connect-by-hostname" and "accept-by-hostname" methods.

Note that these additional advantages provide strong incentives for
operating system vendors and application programmers to adopt the new
stack architecture.  Efficient multi-homing and simplified renumbering
alone provide only weaker incentives, because they have no or very
limited benefits for host and users.

Let me re-emphasize that the proposed new stack architecture may not be
sufficient to fully solve the routing scalability problem.  It will
certainly be a big step, because new multi-homing capabilities and the
reduced renumbering costs will make it easier to use provider-allocated
addresses at the Internet edge.  But some sort of network support may
still be needed to eliminate renumbering altogether.  Moreover, the new
stack architecture could serve as a framework for core ideas from other
host-based solutions.  Take multi-path TCP as an example.  Or the
capability from (host-based) Six/One to rewrite IP addresses in the
network as a trigger for hosts to redirect traffic to a different path.

Anyway, I am writing this down now in said concept paper.  And while I'm
doing this, I'd like to queue up for a presentation slot in Minneapolis.

- Christian



_______________________________________________
rrg mailing list
[email protected]
https://www.irtf.org/mailman/listinfo/rrg

Reply via email to