RE: [PWE3] Generic vs Preferred CW

["Yaakov Stein" <yaakov_s at rad.com>]

> A 1 size fits all approach is just silly, and 
> its no wonder folks will want to deviate from it.
> 
> The analogy for application adaptations into IP would be to 
> say that we don't actually need UDP (with/without RTP for 
> some cases) and TCP, so let's just define 1 adaptation that 
> will fit everything...why not, its exactly the same argument? 
>  The answer here of course is easy to see:
> Applications can be generally grouped into either messages, 
> files or streams, and their functional adaptation 
> requirements are clearly different.....so its pretty obvious 
> a single adaptation will be rather silly as some functional 
> fields will be turned-off for certain application types.

Neil, 

I think you summed up the point rather well,
but I will twist it around a bit.

There is an obvious difference between CO MPLS
and CL IP networks, and thus performance, 
defect handling, etc will all be impacted.

But most characteristics of the adaptation,
(and hence the encapsulation format) depend on
the application (read: client layer) and not on the 
infrastructure (read: server layer).
That's why the different PW types use the PW control
word somewhat differently (e.g. for some the sequence number
is MANDATORY while for others it is OPTIONAL;
some define various flags, etc).

Thus we COULD define different encapsulations,
if we wanted to stress the fact that one server layer is CO
and the other CL, but the differences would only be syntactical,
not semantic. For example, the L2TPv3 PW headers 
and the PWE ones are very different syntactically,
but share the same functionality (e.g. flow demux indicator, 
optional sequencing, various bits to represent the client's
state, etc)

But why SHOULD we use different syntax?
It certainly seems like a useful simplification,
just as standardized drivers or APIs are used in software engineering
to allow the same application to run over different hardware platforms, 
which admittedly may have completely different capabilities.


Y(J)S



_______________________________________________
pwe3 mailing list
pwe3 at ietf.org
https://www1.ietf.org/mailman/listinfo/pwe3