Re: Extending shim6 to use multiple locator pairs simultaneously

[Erik Nordmark <erik.nordmark@Sun.COM>]

john.loughney@nokia.com wrote:
> Eric,
>
> For SCTP + SHIM (which Marcelo & I had a hallway chat about - and maybe
> should write-up), there could be confusion about what the 'path' is
> when combined.  I think the SHIM API should have a mechanism to signal
> up to the transport layer that the SHIM layer has re-homed so that the
> transport layer could decide what should be done - if it thinks
> slow-start
> should be run, etc. 

I don't understand how your comment relates to Scott's idea to run 
multiple paths in parallel. You seem to be talking about shim and sctp 
interaction, which wasn't what I and Scott was discussing.

If you are talking about how SCTP and SHIM6 would interact, I think we 
want to start from the point of SCTP setting the DONTSHIM option, so 
that the shim and sctp aren't used for the same communication.

Past that, it makes sense to study if there are more interesting ways 
for sctp and the shim to share the multihoming work. Potentially this 
can provide a more secure way that sctp's addip functionality. But I 
don't have a good idea on how such a split should be done.
Anybody interested in looking into this?

   Erik