Lambda labels - next steps

["Adrian Farrel" <adrian@olddog.co.uk>]

Hi,

I think some progress got made.

There seems to be some agreement that there may be a need to advertise
lambda availability, use lambda availability in path computation, and
apply explicit label control through the network. These requirements still
need to be firmed up a bit (for example, are we aiming for interworking
between vendors?).

Some say that such requirements imply a global semantic for the lambda
label. Others say that it is sufficient to use local semantics applying
"translation" at each hop. This item needs debate and resolution since it
amounts to "are current protocol elements sufficient *and* optimal for
meeting the requirements?"

Issues of bundling need to be documented, but don't currently appear to
create requirements for new protocol extensions.

John and Payam point out that the wavelength on its own may not be enough
to guarantee end-to-end interworking in optical systems. Since this
interworking is precisely the requirement, we need to set out what
information is actually required.

As Sidney has said, the actual solution comes later (although, since we
are engineers, who can be surprised that we are trying out ideas as we go
along).

My guess, FWIW, is that if you know the ITU grid identifier and the entry
in the grid to which the lambda corresponds then you know everything that
you need to know.

We have an agenda slot for this work in Vancouver. It would be helpful if
the main objective could be to establish consensus on exactly what the
requirements are, and that we cannot solve them using existing protocol
elements. If we get that far, the rest is easy.

Thanks,
Adrian