Re: [mpls] draft-bishnoi-mpls-mldp-opaque-types-01.txt and figuring out all these types ..

["BISHNOI Sandeep" <Sandeep.Bishnoi at alcatel-lucent.com>]

Luca,

Interface ID TLV from RFC3472 and 3471 are unrelated. Interface ID TLV
is for control and data plane association in GMPLS.

P2MP LDP LSP identifier that 'draft-martini-pwe3-p2mp-pw-01' refers is
actually defined in 'draft-ietf-mpls-ldp-p2mp-08'. 

'draft-ietf-l3vpn-2547bis-mcast-bgp-08' is just referring to P2MP LDP
LSP identifier defined in 'draft-ietf-mpls-ldp-p2mp-08'.

There is only single 'generic ID space' defined in
'draft-ietf-mpls-ldp-p2mp-08' and most related drafts refer to it. Our
proposal is to have multiple P2MP LDP LSP identifiers for each
application (optional - unique for each application that dynamically
request mLDP LSP setup). Recommendation is to have separate ID for
static mLDP P2MP LSPs that would avoid overlap among multiple
applications. This would also avoid carving out single space by multiple
vendors that may not be compatible with each other.

Thanks,
Sandeep

-----Original Message-----
From: Luca Martini [mailto:lmartini at cisco.com] 
Sent: Monday, November 09, 2009 2:27 AM
To: BISHNOI Sandeep
Cc: mpls at ietf.org
Subject: draft-bishnoi-mpls-mldp-opaque-types-01.txt and figuring out
all these types ..

Sandeep,

As I mentioned in the meting here are the many versions of these things
:

sec 5 in draft-ietf-mpls-ldp-upstream-04.txt quotes an RFC3472 TLV which
is defined in sec
8.1.1. in RFC3472. This rfc defines an interface ID TLV , and 3471
defines an interface ID.

In draft-jounay-niger-pwe3-source-initiated-p2mp-pw-03.txt it quotes
draft-ietf-mpls-ldp-upstream-04.txt .

Then we have draft-martini-pwe3-p2mp-pw-01.txt , which is the merge of
the draft-Journay-niger , and draft-martini-pwe3-p2mp-pw-00.txt ,and I
have yet another definition of the LSP identifier , which comes from the
L3VPN draft-ietf-l3vpn-2547bis-mcast-bgp-08.txt document. ( this is
going to be changed in the next version)

ok, Now I'm really confused !
:-)

The problem:
We are looking at somehow identifying an P2MP or M2MP LSP.

I would like to state that the specific encoding is trivial, as long as
we can get a unique ID we can all agree on.

I do not agree that we MUST not share the LSP. If there are applications
that are all going to the same locations, I do not see why we cannot
share the tree.

Why are we attempting to centralize the definition of these LDP LSP
identifiers ?

Maybe we just need a simple document defining an IANA registry, and
rules to get assignments from it .
This is very similar to the PW AII definitions, and history shows us
that a simple registry is a good way to solve this problem.

Luca