[Nml-wg] multi-layer inter-domain path finding (was: Meeting notes Sep.24)

[Freek Dijkstra]

Paola Grosso wrote:

> - we need an usecase with multi-layer inter-domain path finding. This
> might (or not) prove John's point we _do_ need transitional links;

Well, there is a multi-layer inter-domain path finding use case. (sorry,
but yes, I'm plugging my thesis once again. Sorry...).

It does path finding without transitional links.

The real question of course is what algorithm is used for path finding.
I defined one algorithm; a colleague, Fernando Kuipers Delft University,
 proposed another one. Both algorithms were exact. Fernando's algorithm
required all topology information in advance, and in practice, so did
mine. Of course, that makes neither algorithm acceptable, so I'm pretty
sure inter-domain path finding will end up with a heuristic algorithm.

I'm saying that the use case is pointless as long as it is not defined
what algorithm is to be used.

There is an algorithm that does not require transitional links, and it
is likely that there is an algorithm that does require transitional links.

Since defining an algorithm is well out of scope of NML, whoever does
make these algorithms should suggest the use case.

Regards,
Freek