I had a hard time with classIDs issue as presented by Tom. I am somewhat neutral on the idea of using a two levels to send termination descriptions. I am not sure that this will be more efficient to implement in the real world, but that was not my objection.
My objection was to the idea of standardizing these classIDs. If the rationale for these is the efficient transmitting of termination descriptions, then they should be modeled to the set of descriptions that will be in use by the MG in question. Since there is a strong relationship between MG and MGC (or between MGCs), these can be sent once at either the time of relationship formation or of classID creation, and get all the efficiency benefits that are proposed. Does anyone believe that there will be any restrictions on creating specific classIDs to meet specific MG needs? If not, that seems sufficient to handle the purpose being proposed with no loss in generality.
Every time we put something in a standard, we create a legacy issue that all future systems are forced to deal with. I do not see any specific advantage in codifying what these classIDs shall contain.
jerry