Issues for discussion from sming wg meeting

["Durham, David" <david.durham@intel.com>]

Greetings, 

I just wanted to try to summarize some of the main discussion points as
gleaned from the voluminous meeting minutes (sorry about that). 

Technical items for discussion as distilled from the 49th IETF sming meeting
minutes:

* Explicit oid mappings: Suggestion to use explicit oid mappings in the SNMP
table declaration to avoid versioning problems when using multiple
implements. There seems to be a problem with implicit columnar numbering for
tables in the case that a table implements classA, classB, and classC (in
that order) and then classB is updated with new attributes (what happens to
classC's implicit numbering?). One solution is to explicitly assign the oid
suffix for each column in the table. 

* Legacy Naming: Why is there a need for SMIv2 ascii names in sming's table
declaration? Several advocated removing the legacy naming convention for
MIBs altogether. 

* Bags and Such: Sets, sequences and discriminated unions. What is their
value and how can they be added (and supported w/in SNMP today)? Seems that
these additional language primitives need to be written up as a specific
proposal before we can proceed.

* Relations: Generic/Specific Pointers, associations, etc. How generic can
we be? Again, a formal proposal for additional ways to express relationships
would be needed to make progress.

* Methods: Is there a need for methods/signatures/formalized
operations/interface definitions. There seems to be support for them if they
somehow formalize and simplify the operational aspects of SNMP... But, there
needs to be a formal proposal before the WG can evaluate the pros and cons
(I believe someone already volunteered to do this). Choices:
	 1. No methods.
	 2. Methods in Data Definition Language for SMIng.
	 3. Methods in DDL for SMIng and mechanisms to map operations in
SMIv2.
	 4. Formal methods in DDL, and convince the SNMP WG to directly
support them.
	 5. Indirectly model methods in the DDL with only minimal additions
to sming primitives.

* Indexing: INDEX, EXPANDS, EXTENDS, REORDERS and additional indexing
semantics. What are the additional indexing mechanisms that need to be
formalized. A proposal would be useful here.

* ABNF Extensions: ABNF statement for language extensions... There was some
discussion as to whether or not the sming should allow the ABNF as it could
potentially be abused. It is currently used to formally describe extensions
to the language.

* Selective Class Implementation: Partial implementation of a class in a
SNMP table. Should this be disallowed? How are backwards-compatible mappings
done otherwise? How would versioning be handled otherwise?

* 7-bit ASCII: Why have a 7-bit ASCII restriction?


Please let me know if I missed (or misrepresented) any technical points that
we need to discuss.

Cheers,
-Dave