Why isn't there an LSC in GMPLS?

[Kshitij Kumar <kshitij@rogers.com>]

Folks,

I am curious. Please humour my curiousity.

There are PSCs, L2SCs, etc. in GMPLS. 

Take the case of p2p links, GigE for instance.

A PSC (for instance an IP router capable of MPLS forwarding) needs to look at the MAC address, then look at the label and make a forwarding decision, changing the MAC SA/DA in the process.

An L2SC (for instance an Ethernet switch capable of doing MPLS) checks the MAC destination and presumably make a forwarding decision based on either the MAC or the label. If it is a pure L2SC, I assume it makes the decision based on the label and should not change the MAC SA/DA. Why does an L2SC look at the L2 header? What's the relationship between the L2 header vs. Label for an L2SC? This is never explained in the GMPLS drafts.

And, there is no "Label Switch Capable" (LSC) device that I know of, or is there?

An LSC would be a device that would simply switch based on labels and labels only. It would have interfaces that would be P2P - for instance GE or FE p2p links, and would not spend cycles looking at the L2 header for making forwarding decisions. 

The LSC would be a "real" GMPLS device - one that is "Generalized" not just at the control plane, but also at the data plane. (Of course, one can't generalize the implementation, but you get my point).

It's big advantage could be that it could be transparent to any L2 bridging or L3 routing - it would work purely based on MPLS tunnels. (Will need some routing for GMPLS signaling, but that's it).

Has such a concept been discussed and if so, what happened? If not, why not?

Thanks,

Kshitij.
--
My cure for procrastination: 
"I wasted time, and now doth time waste me."
King Richard-II, Shakespeare.