[Tsvwg] architecture for transport layer mobility

At the San Diego meeting last August, Allison closed the meeting by
talking very broadly about hosts having multiple addresses, and how this
affects the transport layer [1].  One of the potential uses is for
handling host mobility at the transport layer.  This has been done
experimentally several times in several different transport protocols,
but not within any specific unified architecture.

For example, over the years there have been mobile SCTP, TCP, and DCCP
proposals that all have used different code for supporting movement
detection (finding out when the local host has moved) and doing location
management (providing some way for the mobile node to be reached for new
connections).  These are two common problems for all transport layer
mobility protocols, which happen to be independent of the particular
transport protocol used.  It makes sense to define a common framework to
provide this functionality, so that individual transports don't have to
worry about it.

As a first step, we've begun exploring a simple architecture to provide
these two services using existing standards (eg neighbor discovery,
dynamic DNS, etc), and described our architecture in:

http://www.ietf.org/internet-drafts/draft-eddy-tlmarch-00.txt

We would appreciate any feedback from the community on this, and would
welcome the opportunity to speak breifly in Washington if the group is
interested.  We feel that while there is little new here, providing the
common framework for transport protocols to easily develop mobility
support with little replication of effort, is an important step.
Additionally, this draft attempts to provide a detailed comparison of
the architectural differences between transport layer mobility and
Mobile IP, the predominant mobility protocol which operates at a
completely different layer and has different deployment and stack-design
considerations.

-Wes

[1] See meeting notes at http://www.ietf.org/proceedings/04aug/261.htm