[sacm] Endpoint ID Design Team Notes: 2015-10-09

["Haynes, Dan" <[e-mail]>]

Hi Everyone,

Here are the notes from the 10/09 Endpoint ID Design Team Meeting (EID-DT).

Attendees

*         Ron Colvin

*         Henk Birkholz

*         Ira McDonald

*         Danny Haynes

Note Well

Agenda
*       Further discuss the triple example with updates
*       Discuss how it could be integrated into the Information Model
*       Representation of constructs in the information model
*       Open issues from the last virtual interim meeting

Materials

*         Slides from meeting (see attached)

Notes

*         First, the triple example was discussed.  In doing so, it was suggested that requiring the explicit expansion of subject, predicate, object is cumbersome and it may be better to define some requirements for attribute-value-pair (AVP) relationships (i.e. principles for the information model).  Two such requirements might be as follows.  (1) If you define an AVP, you must define another AVP that it relates to.  For example, you can't just say IP addresses are important, you must show how it relates to something else (i.e. a network interface has an IP address).  (2) Define basic relationships that data models must support such as an is-a relationship (taxonomical) and a has-a relationship (yang).  From there, you just require that data models support these requirements and after that they are free to do what they choose.  This would allow both triples and yang to be used in SACM.



*         There was a question about how quickly the model would be able to be updated to support new objects over time.  It was suggested that you may be able to set something up like an IANA registry for data models.  A concern was also raised that the information model must be represented in a formal syntax of some sort.  It was also noted that we want to enable organizations to develop their own extensions, but in doing so, we will want them to map their extensions back to the information model.  With respect to IANA, using MIBs as an example, they want documents to point at.  Without Internet-Drafts that get pointed to and is represented as a yang model and is explicit as objects with syntax.  It was also mentioned that extensions aren't particularly useful if you do not give people an precise schema.  As an example, it was explained how we might want to look at the yang models for interface (https://tools.ietf.org/rfc/rfc7223.txt) and system management (https://tools.ietf.org/id/draft-ietf-netmod-system-mgmt-16.txt).  There was also interest in seeing a small example of a yang module extension of the information model.  It was also noted that some people like UML as a way to model data.  Another thing to look at is DMTF CIM which has a lot of models and is being used by many operating system vendors.  It has a good model for software and might provide a good starting point for attributes that we may want to consider.



*         Open issues from the last virtual interim meeting were not discussed.


Action Items

*         Add tracker for CIM.  Done (https://github.com/sacmwg/draft-ietf-sacm-information-model/issues/31).  (Danny Haynes)

Please let me know if I missed anything, got anything wrong, or if you have any questions.

Thanks,

Danny