Hi guys- Here are some slides presenting a general concept of how I think we can a) provide a more comprehensive NSI topology model that provides more internal structure, and b) gracefully dovetails with existing NML dataplane constructs. Look at the constructs and the concept, we can adjust the syntax appropriately. I am at the OnVector workshop and present shortly after the call, so I may not be able to stay on today's call very long. Thanks! Jerry
Hi Jerry and others, Looking at your slide, I see two concepts, and I'm not sure if they're the same or not. The first concept is the hierarchical topology: describing a topology in more and more (or less and less) detail. The second concept is the alias: using a pointer to point from a generic port or link ("the link between Nordunet and MAN LAN") to a more concrete network element ("the link between Ciena 4000, intf 2/1 in CPH001a and the Foundry 1234 intf 8/0 in NY002a"). Both concepts seem to deal with abstraction and can help distinguishing between a _functional description_ of a network, and how those function are _implemented_. However, the first concept is mostly about aggregation, while the second is about versioning. While I see aggregation as some "geographic" thing (how far do you zoom in area), the version is a "time" thing (from when to when is this conceptual link in place, and when is it ported to a different device). I think NML can support either concept, but I wonder if we should support both concepts, or if supporting one is good enough, or perhaps both concepts are the same after all. Regards, Freek
Jeroen wrote in the notes:
We had a short discussion on whether NSI descriptions and physical topology descriptions are different things. One is something like a service, the other is the actual instantiation of it. This is similar to identifiers for books, stories, translations and physical books. Freek remembers that libraries have a solution for identifiers on those things, he will try to get a link to that and post it.
My feeble attempt to translate library versioning concepts to IT: * work - Eg. "Gone with the Wind" or "Browne natural language corpus" * expression - Swedish translation of the work - software implementation * manifestation - hard cover of the expression - Excel sheet or XML file * item - actual book on a shelf - copy on a disk Anyway, in NML we currently have a "Lifetime" for each network element (it's in the diagram). I wonder if the aliases Jerry mentioned in his slides are a better way to handle versioning. (Assuming this is all different from topology aggregation.) Freek
participants (2)
-
Freek Dijkstra -
Jerry Sobieski