The MPLS WG Archive

[David Charlap <David.Charlap@marconi.com>]

Manikantan Srinivasan wrote:
 >
 > 3) Can you explain why do we need both TunnelID and LSPID in the LSP
 >    sub object? The Ingress, Egress Addresses and the TunnelID is
 >    sufficient to uniquely identify a tunnel.

It is not sufficient.  In order to uniquely identify an RSVP-TE LSP, you
need no fewer than five values:
	Tunnel endpoint (egress) address (in SESSION object)
	Tunnel ID (in SESSION object)
	Extended Tunnel ID (in SESSION object)
	Sender (ingress) address (in SENDER_TEMPLATE/FILTER_SPEC object)
	LSP ID (in SENDER_TEMPLATE/FILTER_SPEC object)

Any implementation that makes assumptions about these values is broken.
The existing MPLS MIBs (LSR and TE) do not contain enough information
to cover all of RSVP-TE's capabilities.  For instance, the extended
tunnel ID does not have to match the sender address (it could also be
zero).  And one tunnel (defined by the three values in the SESSION
object) may contain multiple LSPs, each with its own separate LSP ID.

If you don't permit one tunnel to have separate LSPs, then you
effectively eliminate SE style and all of the useful features that it
offers.  You also drasticly reduce the maximum number of LSPs that can
be created between two endpoints.

-- David