NML Session 1:

Agenda:
 - Update on ITU and TMForum
 - Layer Definition
 - Path/Link
 - Crossconnect


Freek reviews progress on the schema draft and the open tasks.

Freek then provides a summary of the network modelling efforts in the ITU and TMForum (slightly hampered by their IPR policies).
It is generally agreed that it is interesting to know what the ITU and TMForum are up to, but that we are mostly working in parallel. Ideas could be exchanged, however since ITU and TMForum are generally closed this is hard.

Jeroen and Freek have revisited the definition of Layer and have come up with a new proposal:
	Layer: A type of encoding, so that a source Port and sink Port of a common layer can be associated together.

The question comes up whether a Layer is a collection or an attribute of objects. The intention is to make it a collection.

Is a single port on a single layer? Yes.

Physical and logical ports can be grouped together, also with things like VLANs.

There is rough consensus on restricting layers to have only one label, with the addition that it should be described as a "single type of label".

A Label is something that is used to identify a channel in a multiplexing adaptation.

The definition of Layer is currently okay, but improvements are always welcome!

G.800 splits the adaptation and termination concepts and uses the concept of 'trail'. The group does not see a reason to add this to the NML model.

Vangelis proposed the word "Connector" as a better term for Link. We currently don't see a reason to add this.

Martin suggests that Links can be between Ports, but also between Domains or other Network Objects. We need to examine this use-case closer.

Martin presents his and Chins work on the definitions of Path and Links.

A Path is an ordered collection of Network Elements.
There is a discussion on putting restrictions on the collection:
- Should it be a collection of just Links? or just Ports?
- Does it have to be connected?

After some discussion it is decided that this will heavily depend on the context. We agree that it is best to leave the definition flexible and allow applications to put in their own restrictions on the Path format.

A Link can be used to abstract away the details of a Path.

Session 2

Freek presents two use-cases for Link/Path/Segments with a global lightpath crossing several domains. Some of the domains use their own identifiers for their section of the lightpath.
Martin suggests that there should be relations between the domain identifiers and the global identifier.

It is hard to require a global identifier. GLIF is using a scheme, but NML cannot impose a global ID scheme.

Martin will look into the name and definitions of the relations between local and global identifiers.
Freek and Inder will look into the global unique identifier issues.

Freek presents his and Jerry Sobieski's work on crossconnects.
There are three different functions involved:
- Transport function - moving data from A to B
- Transform function - changing data from one format to another
- Transfer function - a combination of transport and transform

Do the functions descibe possible or existing scenarios? Unclear.

The current scheme is using unidirectional ports and crossconnects. Is there a relation between input and output ports?

John MacAuley: if you do not define any relations between the ports, you can run into problems when provisioning.

Martin and Gigi will look at the implications of unidirectional topologies.

Guy Roberts: is a switching function tied to physical things?
This should be possible to define, since there are machines that have capabilities depending on line-cards present in the machine.

John MacAuley: Does NML provide enough information for protection?
It is possible to add attributes to Network Objects to define things like Shared-Risk Link Groups, etc.
John MacAuley will write up a use-case for protection.

Jerry and Freek will look at their definition of multiplexing again.

Adaptation can happen in a line-card, so it should not be part of the switching-matrix.

We will setup bi-weekly teleconferences to keep work going.
Jeroen will setup a Doodle to find a timeslot for a teleconference.

Vangelis wonders whether it is possible already to use the NML schema to describe topologies for the Dynamic Gole demonstration.
It seems that the current schema is nearing completion, so this should provide some valuable feedback.

Vangelis and Jeroen will look into applying the schema to this demonstration.

Action Points:
- Martin will look into the name and definitions of the relations between local and global identifiers.
- Freek and Inder will look into the global unique identifier issues.
- Martin and Gigi will look at the implications of unidirectional topologies.
- John MacAuley will write up a use-case for protection.
- Jerry and Freek will look at their definition of multiplexing again.
- Jeroen will setup a Doodle to find a timeslot for a teleconference.
- Vangelis and Jeroen will look into applying the schema to the Dynamic Gole demonstration.