[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: thoughts on draft-vasseur-ccamp-inter-area-as-te-00.txt



Hi Dimitri,

At 04:24 PM 3/2/2004 +0000, Dimitri Papadimitriou wrote:
hi jp,

see in-line:
>
> Hi Dimitri,
>
> At 02:45 AM 3/2/2004 +0000, Dimitri Papadimitriou wrote:
> >hi arthi, jp, all,
> >
> >reading draft-vasseur-ccamp-inter-area-as-te-00.txt
> >the following came into my mind, why is there a need
> >to signal the "signaling method": contiguous versus
> >stitching vs nesting ?
> >
> >the reason provided in the document about "control"
> >is unclear, is the sensitivity of the carried traffic
> >dependent of the signaling method ?
> >
> >"For the sake of illustration, a Head-end LSR, may
> >desire to prevent stitching or nesting for a traffic
> >sensitive inter-area/AS TE LSPs that require a path
> >control on the head-end LSR."
> >
> >it seems for me that "signaling the signaling method"
> >introduces here in addition to the protocol issue(s),
> >policy issues:
> >
> >"Ox01: Contiguous LSP [...]
> >
> >When set, this indicates that a boundary LSR MUST
> >not perform any stitching or nesting on the TE LSP
> >and the TE LSP MUST be routed as any other TE LSP
> >(it must be contiguous end to end). [...]
> >
> >A mid-point LSR not supporting contiguous TE LSP
> >MUST send a Path Error message upstream with error
> >sub-code=17 Contiguous LSP type not supported."
>
> because when crossing multiple ASes this allows the head-end signalling a
> contiguous LSP to make sure that it will not be stitched in some downstream
> domain hence keeping some strict control about the reoptimization. This has
> to be done on a per-LSP basis, since the same head-end may not impose any
> such requirement for other inter-area/AS LSPs.


not sure i understand that what you want to achieve requires
the method that you are proposing here: in the hypothetical
case of multi-as/area reoptimization, what you say is "i want
to keep head-end control for the re-optimization of a multi-
area/asn LSP"

so here, why don't you use a flag saying "head end control
of re-optimization" meaning we define a method such that the
head-end decides its willingness to keep control of the re-
optimization occurrence and upon notification could decide
about when this has to be committed and this *independently*
of the LSP provisioning method for this LSP

in the meanwhile this would allow to be more efficient about
the way each ingress boundary LSR decides about the method to
deliver the incoming request constraints

so, in brief define a method that would avoid constraining the
provisioning method from the head-end as/area perspective to
its downstream as/area for an hypothetical re-optmization that
is in any case under the control of the local domain, and this
in particular the local as)

Well, no, the intention here is really to offer to the head-end the ability to specify (just as any other LSP attributes) the signalling method that must be used in other downstream area/AS (note that this is clearly a SP requirement). Indeed, the reoptimization is just one component but there are also other aspects related to the signalling method in use. Hence, the aim of this bit is indeed to specify the signalling type of the inter-area/AS TE LSP.


thanks for your comments


ps: unfortunately you're not attending this ietf meeting so
that we can't discuss this important issue during the ccamp
face-to-face meeting

JP.


thanks,
- dimitri.
> Thanks
>
> JP.
>
> >thanks,
> >- dimitri.
>
>
>