Re: administrivia (on avoiding injury)

[Francis Dupont <Francis.Dupont@enst-bretagne.fr>]

 In your previous mail you wrote:

   > The kind of hackery involved in address selection by the host stack, sure.
   > We've always known there would be hackery there. There's no reason
   > that the restart must use the same address pair as the aborted session.
   
   How do you accomplish getting an unmodifyed station to change the source
   address it uses?  
   
=> today the obvious solution is to deprecate the prefix of the bad address
but some kind of preference between prefixes should be standardized
ASAP in order to add more flexibility.

   Walking through a DNS round-robin would solve it for needing to use a
   diffent destination (providing the hosts actually tries other IP addresses
   after a link fails),

=> on a standard BSD system we ("INRIA" stack team) found only
one application (lpr) which tries only the first address returned by
the DNS. So this scheme works well!

   but I can't figure out source..
   
=> hack source address selection (as proposed in a phase 2 about this
in an old IETF) ?

Regards

Francis.Dupont@enst-bretagne.fr