RE: New charter

[<neil.2.harrison@bt.com>]

Hi Scott....good remarks, a couple of observations below:

Scott wrote 27 November 2004 13:13

> On Fri, Nov 26, 2004 02:37:37PM +0100, Loa Andersson allegedly wrote:
> > All,
> > 
> > there seems to be a majority for doing L1VPN in
> > ccamp, I don't particular object, but remember the arguments when we

> > placed the L3VPN and L2VPN in the Internet Area. The main argument 
> > was that the Internet Area deals with end-2-end services
> > over the Interent, thus VPNs belongs in the
> > Internet Area.
> > 
> > If we decides that L1VPN don't go together with
> > L2/L3VPNs in the Internet Area we should at least
> > have some motivation for this. What is it?
> > Is L1VPN not an end-2-end services or is it not
> > an Internet Service?
> > 
> > /Loa
> 
> Is it that L2 and L3 VPNs are services on top of the
> infrastructure provided by the network, but L1VPNs can 
> allocate the infrastructure resources themselves?
NH=> I think we as an industry need to start to decoupling ourselves
from this L1/2/3 classification as it has no value.  Any network that
has addressing and routing functional components is a layer network (in
the obsolete/restricted OSI L3 meaning of 'network layer').  IP is a
layer network, and it belongs to the cl-ps mode...which is great.  FR,
ATM and MPLS are layer networks, and these all belong to the co-ps
mode...which is also great.  PDH, SDH and OTNs are also layer networks,
and these all belong to the co-cs mode...and this too is great.  But the
3 network modes are different and trying to apply a 1 size fits all
solution simply won't work.  Indeed, it is the modal differences that
are actually THE important bits, this is where the real value
lies.....so we should not knock this.  If folks want to classify layer
networks they should use these 3 modes and the G.805/809 methods of
describing them.  This is the only classification that makes any logical
and practical sense.


> L1VPNs 
> aren't (necessarily) "over" anything?
NH=> To be accurate we can put any X (layer network/mode) over any Y
(layer network/mode).  Whether its makes real sense to do this (and/or
how many times such transitions should be allowed in some global
reference model) is another question.

>  In G.805-speak, unlike 
> L2 and L3 VPNs, they aren't necessarily path layer networks.
NH=> Scott...when we create a layer network (any) the link connections
in that layer network are invariably formed/supported by trails (which
are end-end entities) in some server layer network.  In turn, the
link-connectines of these server layer trails are supported by even
lower level layer network trails.  This is a recursive behaviour until
we hit the duct.
 
> It is an end-to-end service, but not necessarily an Internet 
> service (the control of it is)?
NH=> I think some folks have a weird notion of what end-end actually
means. In a layer network end-end means between access points (which are
closely bound to trail termination points).  So one guys end-end network
is some other guys link-connection, ie a 'network builder service' for
example.  End-end is in the eye of the layer network beholder when we
talk of layer networks.  There are 2 end-stops of layer network
recursion however...the duct (obviously) and the customer end-system
applications (like voice, Email, whatever).  But between the application
and the duct there can, and are, many layer networks.

regards, Neil