[Nsi-wg] STP

[Jeroen van der Ham]

Hello,

Summarising the presentation and yesterdays discussion, the proposal Tomhohiro is making here is in line with Jerry's earlier proposal. 

The underlying model is that for every network-port and label combination a new STP is created. Should a network not support VLAN retagging, then that network is represented as 4000 different networks, each with the respective STPs.

The slides show a convenient way of aggregating STPs based on their format. The syntax also allows domains to present rules so that they don't have to manually create 4000 different networks.


My opinion is still that this model is trying to muffle away the complexity of label assignment. I believe that labels should be treated as first-class objects, and become part of the setup phase of a connection. Having a *typed* label makes it almost trivial to move to a multi-layer model later on.

A multi-layer model without explicit labels means that the representation will balloon to even more STPs, and network representations, with more complicated rules for each domain to manage the inter-dependency of multi-layered circuits.

Jeroen.