Re: Transport level multihoming

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

Greg;

> Has anyone here had any discussion about the potential for transport level
> multihoming (like in SCTP, RFC2960) to replace the current practice 
> globally advertised multihoming on the IPv6 Internet?

> Has a potential policy of 'No global IP multihoming of short prefixes;
> Leave multihoming to the end nodes' for IPv6 been discussed here? 

Yes.

It is possible to add a TCP option and let new TCP handles multihoming
while retaining interoperability to old TCP implementations.

> If the transport pushes multihoming to the end node and IP layer
> multihoming not being propagated beyond direct peers, it would seem that
> such a system could significantly further the scalability of the Internet,

Yes, some applications may be satisfied by a standard multihoming
subroutine, library or system call (for the Internet, there actually
is no point to distinguish "transport"and "application" in a end
system, because it is not visible to the network protocol).

However, transport layer is not a proper place to stop pushing.

TCP based applications need minor changes and UDP based applications
may need more changes.

Thus, the scheme is called "end to end multihoming".

							Masataka Ohta