Hi Freek/All; As in my response on your/Jerry's other proposals, some example are required to highlight the positive and negative aspects of each approach to accurately gauge the argument why there needs to be a change. Thanks; -jason On 3/4/12 8:14 AM, thus spake Freek Dijkstra:

Each NML object has a "Lifetime" attribute which specifies -well- when an NML object is valid.

I see two problem with that, namely that it is often not known in advance when the lifetime of an object terminate, and that describing the lifetime of each object is not very efficient.

Jerry's idea about versioning and topology updates made me reconsider the current proposal.

Jerry's idea is to simply change a whole topology description (or subtopology description) when it is changed. This seems more useful than to change the lifetime of each individual network object.

Hence I propose to get rid of the Lifetime object and replace it with a version attribute, which is a timestamp.

The version attribute should be used for topologies (or for any network object if the WG prefers). The meaning of the version attribute is that the topology with the latest timestamp (before now -- see below) describes the current active topology. This is much like the version string in DNS SOA records. A topology can be end-of-life'd by describing the topology without any ingress or egress Ports, and giving it a new timestamp.

One of the great features of the "Lifetime" attribute is that it would allow description of future topologies, and hence advance reservations. This is why the "before now" in the above description comes in. It would allow one to describe a topology with a timestamp that lies in the future. The meaning would be that that topology would not be valid until that time is reached, hence allowing the description of future topologies as well.

This seems much easier than the current proposal, and just as flexible.

Can we decide to replace Lifetime with version?

Regards, Freek