RE: Fwd: Minutes / Notes

["Tony Li" <Tony.Li@procket.com>]

I think a tax that is of a bigger concern is that we need to affect most
of the predominant implementations out there.  Once we get past that
hurdle, having a host that knows that it's multi-homed or it's
corresponding
with a multi-homed host is a small additional tax with substantial
benefits.

Tony


|    -----Original Message-----
|    From: Christian Huitema [mailto:huitema@windows.microsoft.com] 
|    Sent: Sunday, July 20, 2003 10:02 PM
|    To: Tony Li; J. Noel Chiappa; multi6@ops.ietf.org
|    Cc: jnc@ginger.lcs.mit.edu
|    Subject: RE: Fwd: Minutes / Notes
|    
|    
|    > Now, if you don't want to play, I understand, but doing something
|    > to fix the current situation is absolutely essential.  We cannot
|    > have scalable routing with PI addresses.  These are also a tax on
|    > everyone and a much higher tax where we can ill afford it.
|    
|    I definitely believe that we need a solution for 
|    multi-homed sites. I just want to make sure that whatever 
|    tax we create is only paid by those who benefit from 
|    multi-homing, i.e. multi-homed hosts and their 
|    correspondents. Moreover, I submit that the tax should 
|    only be paid py those applications that actually benefit 
|    from TCP resiliency, i.e. those whose "mean connection 
|    time" is not small compared to the "mean time between 
|    rehoming events". Even so, applications that have their 
|    own identification and reconnection mechanisms should not 
|    have to pay the identifier tax.
|    
|    -- Christian Huitema
|    
|    
|