RE: [PWE3] PWE Service interworking

[Keith Knightson <kgk@igs.net>]

At 11:39 PM +0100 7/6/03, neil.2.harrison@bt.com wrote:
> Tom,
>
> If you understand G.805/809 (and therein lies the major problem) the meaning
> of 'end-end' is obvious, ie it's between the trail termination points in
> whatever layer network is considered.  Forget the wrong/misleading OSI
> model......there is more to networking than 'one' network layer.
>
> regards, Neil
Neil,

Oh, please.....now you are going too far....even for your high 
horses.......  :-)

And, if you were to understand IS 8648 and IS 7498  (and therein lies 
the major problem),  you would know about "sub-layers" and 
"relays"..... and how more than two entities can be distributed 
within a given layer/sub-layer to overcome the discontinuities due to 
protocol differences.

You said "There is more to networking than 'one' network layer". 
Yes, and they were called sub-layers, within the named layer in 
question. Any layer can be sub-divided into sub-layers, this is clear 
from IS 7498/X.200 and is particularly described in IS 8648.   You 
are confusing the general concept of layering  (which includes 
sub-layering) with specific religious labelling and packaging of a 
stack for a particular open system design formulated in a bygone era 
of X.25.

You say the meaning of 'end-end' is obvious.. This is clearly not the 
case....particularly since G.805 is actually wrong in this respect.

For example, with service interworking, the "end" points are not 
coincident with your so-called trails or protocol termination points. 
Service mapping implies conveying the *semantics* across a relaying 
point, i,e. beyond the point of protocol termination. Let's just 
suppose for argument's sake that some effect is common to both the 
protocols being interworked and can be mapped to provide an 
"end-to-end" effect.  It is clear that the intermediate points simply 
convey this effect by mapping. Therefore the trail termination points 
of the protocols on each side of the relay cannot be considered to be 
ends for the communication in that layer, despite the fact that each 
of the protocol is terminated at some intermediate point.  There are 
a number of peer entitities operating at the same "level" operating 
co-operatively and in tandem to provide ends beyond themselves.  Nor 
can you consider that there is higher layer trail to rperesent the 
e2e because there is no addtional protocol layer to terminate such a 
trail.  I.e. G.805 simply does not work in this case.

The whole point about services is defining their semantics, i.e 
beyond the one dimensional issue of what's on the wire, that's just 
syntax. Origination and/or Termination points do not necessarily 
dictate the end point for all communication at that particular peer 
level. They only indicate the presence of generators and receivers of 
a particular format/syntax processing.

The real question to be asked if whether the user be aware that the 
so-called service interworking is taking place as opposed to the 
so-called network interworking for any given protocol he uses between 
him (the CE) and the PE?   Suppose both SIW and NIW are involved for 
for two different paths from given CE with the same CE-PE protocol 
stack at one end, but with NIW at end and SIw at another, because 
conditions at each of the destinations are different.   All this 
should be hidden from the user, but the semantics should be the same. 
Users cannot be expected to know how the semantics might differ for 
every conceivable concoction that might be applied deep in the bowels 
of core networks or at remote access networks. BTW, it is possible to 
map OAM imperatives/notifications across a service interworking unit 
for and within a given peering arrangement.

The use of the term "layer network"in G.805/809 is also erroneous and 
not useful in this respect. Trail end-points are clearly not always 
indicative of "networks" nor "ends".  A protocol does not 
automatically beget a network. The term network also has no useful 
universal or absolute definition is also best avoided. Addtionally, 
G.805 ignores the three party semantics between user-provider-user, a 
vital ingredient in building relays which deploy service interworking 
and understanding of the service being provided. I will ignore the 
issue about the usefulness or otherwise of the term "layer" for the 
moment.

In actuality, protocols have roles. Protocols play a role according 
to context in which they are applied. They also operate between peer 
entities.  That's about all that can be said in terms of useful basic 
principles in the general context (i.e.not OSI context).

I suggest avoiding these "my-architecture/your-architecture" 
rat-holes, and simply discuss how protocols are to be stacked, and 
where, and the semantic effects required between any /all of 
designated points that can be identified. Of course, in this case, I 
agree it would be much simpler if the IETF could join the 21st 
century and  allow the use of something sensible to draw proper 
pictures to illustrate all these cases and issues more precisely, 
pictures being worth thousands of disageable and ambiguous words.

Regards

Keith

BTW it is really simple to define the difference between network 
interworking and service interworking without using the terms layer 
or network or G.805 or any other inappropriate model.
_______________________________________________
pwe3 mailing list
pwe3@ietf.org
https://www1.ietf.org/mailman/listinfo/pwe3