The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Jul> msg00398



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Consensus on compression - draft-ietf-mpls-hdr-comp-00.txt

  • From: neil.2.harrison@bt.com
  • Date: Wed, 26 Jul 2000 11:57:41 +0100
  • Cc: mpls@UU.NET

> > PS The "-simple" approach is a totally new one.  This is 
> > appropriate as 
> > draft-ietf-mpls-hdr-comp- and the 2507/8 are "one-hop" based, while 
> > "-simple" is multi-hop.
> 
> Hmmm.  I think there is plenty of work afoot that
> leans toward treating an LSP as one hop.  Indeed,
> part of the issue with your approach is the fact
> that "one-hop" is a bit of a stretchy notion...
> 
	[Harrison,N,Neil,INC1 X]  I have no question in my mind that an LSP
is '1-hop'.  This is an obvious client/server layer relationship.  In a more
strict architectural sense "a client layer link = server layer
trail".......this is indeed how one builds the topology of a client layer
network from a server layer.  This relationship recurses across all layers,
starting from the 'duct'.  Failure to appreciate (and adhere) to this
fundamental principle will cause all sorts of downstream problems....many of
which I have already observed cropping up in several areas such as DiffServ
and the notion of a single control-plane across all layers (but the details
are not for discussion here).