Re: New version of draft-caviglia-ccamp-pc-and-sc-reqs available on repository

[Dimitri.Papadimitriou@alcatel.be]

hi adrian - see in line




"Adrian Farrel" <adrian@olddog.co.uk>
20/06/2006 02:34
Please respond to "Adrian Farrel"
 
        To:     "Diego Caviglia" <Diego.Caviglia@marconi.com>, Dimitri 
PAPADIMITRIOU/BE/ALCATEL@ALCATEL
        cc:     <ccamp@ops.ietf.org>, "Dino Bramanti" 
<Dino.Bramanti@marconi.com>
        Subject:        Re: New version of 
draft-caviglia-ccamp-pc-and-sc-reqs available on repository


Hi Dimitri,


> the doc enters in a discussion in section 1.1 which is the kernel of 
many
> discussions we have had on this list
>
> from this perspective i'd like to know what you are referring to when
> stating
>
> "A Label Switched Path (LSP) has different semantics depending on the
> plane in which it the term is used"
>                                           ^^^^^^^^^

A bit tricky to determine what is underlined here.

[dp] i meant "semantics"

Anyway, you have highlighted and important issue.

Sometimes, when people say "LSP" they mean the data plane state. That is 
the 
sequence of data plane resources (links, labels, cross-connects) that 
achieve end-to-end data transport.

Sometimes people mean the control plane state created to manage the LSP in 

the data plane.

Sometime people mean the management plane/configuration information for 
the 
LSP in the data plane.

So the I-D needs to be clear what it is talking about.

[dp] agreed, my question is more is there really something different from 
base 
MPLS once we assume that a label can have multiple semantics (label, 
timeslot, 
VLAN, etc.) 

> ps: looking at the terms used for LSP in the MP/CP, the corr. paragraphs
> just speak about representation of the data path

Not sure. The authors will probably comment.

Seems to me that the I-D is talking about maintaining the data plane state 

while exchanging the control state between the control and management 
planes. This could probably be clearer in the draft.

[dp] agreed

Adrian