[Nml-wg] NSI topology

[Freek Dijkstra]

Hello Guy, Inder, Tomohiro,
cc: NML and NSI groups

Last week Roman Łapacz (PSNC) took Jeroen's AutoGOLE network description
and turned it into an XML format. (Thanks Roman!)

In that process, we discovered a few differences how AutoGOLE and NML
describe a network model. The main differences are:

* NML links and ports are unidirectional, AutoGOLE links are
  bidirectional
* NML uses links and ports to describe network connections (a
  proposal to describe only links, not the ports was not followed
  up), AutoGOLE only describes the ports (the links are described
  as connectedTo between the ports, but have no identifier of their
  own)


Of course there are some commonalities:
* Both implementations focus on logical connections, not physical
  connections.
* Both implementations have provisionings to ensure a layer of
  abstraction in case attributes of a link or port change (in NML:
  the identifiers are opaque strings and have no inherent attributes;
  in AutoGOLE and NSI: the distinction between SDP and STP)

In yesterday's NML conference [1], we briefly discusses these
differences and commonalities.


Now, I do not know to what aspect the NSI working group is building on
the AutoGOLE work, and what the current issues are. I don't want to
interrupt the considerable progress the NSI group is making, and
certainly don't want to get in the way of more great demos such as
previous year's AutoGOLE demos by interrupting their implementation
timelines.

I think it would be good for the NML group to:
- educate the NSI working group participants on these differences.
- listen to the NSI group if there are any problems with the topology
  as we specified.

I realise that the above may take some valuable resources of the NML
working group, but I am dedicated to make sure our groups are aligned.


My question to you (as NSI co-chairs) is if you think the above is
useful, and if so, when we can best discuss this.  Would it be useful to
discuss this with the whole NSI group, or a subset?  I propose that we
dedicate a timeslot for this work, and am happy to use the NML break-out
slot for this purpose (if you like it earlier, we can see if we can ask
Joel to re-arrange some slots). We can also use a NSI or NML telephone
call to discuss this.

Part of that work means for NML-WG to provide example topology
descriptions (which can be done within a few days). But also means
making code contributions, which I don't think the NML-WG has the
resources for on short term notice.

What is your recommended way forward?

Regards,
Freek Dijkstra

[1] http://www.ogf.org/pipermail/nml-wg/2012-February/000815.html