Re: Network layer reqt? [was Re: Transport level multihoming]

[Greg Maxwell <gmaxwell@martin.fl.us>]

On Thu, 5 Apr 2001, Masataka Ohta wrote:

> Brian;
> 
> > I think Margaret Wasserman's message captured the requirement. An
> > IPv6 host running RFC 2460 IPv6, and running TCP and UDP applications, 
> > including IPv4 applications running over bump-in-the-stack or 
> > bump-in-the-API, must be able to open *new* TCP and UDP sessions,
> > regardless of which of the site's IPv6 provider links is up or down.
> 
> Wrong.
> 
> TCP is the reliable transport.
> 
> Most TCP applications are so stateful that they can not accept a
> loss of bytes.
> 
> If you insist that TCP may be unreliable upon multihoming,
> you will end up with construting virtual reliable TCP over
> unreliable TCP.

I don't think that was the intention. I think the intention is that your
TCP session may die, or UDP may lose packets but be re-establishable if
there is another provider available.  Much like ATM SVCs must be 
reconnected if a path totally fails.

I think such a solution would be acceptable for TCP support. A future
enhancement or replacement such as SCTP could provide non-connection
losing support and other enhancements.

I think that this is acceptable because with convergence time as it is
today in the IPv4 world, TCP connections are often hung when a link fails.