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

[Brian E Carpenter <brian@hursley.ibm.com>]

Margaret Wasserman wrote:
> 
> >No. But today, connections with TCP/IPv4 and typical IP multi-homing get
> >dropped and data gets lost when the prefered path fails. Traffic is lost
> >until the routing reconverges which is usually long enough to back TCP off
> >into oblivion.
> 
> Is this true?  Can other operators or admins of large sites
> comment on this statement, in particular?

Well, we've had many anecdotes of very long (>100 seconds) BGP 
convergence times- Abha reported on some actual *facts* in the 
IETF plenary- see http://www.merit.edu/ipma/. TCP doesn't
react well.

> 
> If this is true for a preponderance of the IPv4 multi-
> homing solutions, then I would propose that it is _not_ a
> requirement that a scalable IPv6 multi-homing solution
> provide a mechanism to maintain TCP connections across
> the failure of an Internet connection point.
> 
> It would only be a requirement that TCP hosts can establish
> a new TCP connection once the routing tables have converged
> to reflect the outage.
> 
> There are several ways in which TCP could be modified to
> improve this situation, but we could consider those
> mechanisms in a different group -- maybe the Transport
> Area WG.
> 
> Who agrees?

Me
  Brian