Re: #12

[Juergen Schoenwaelder <schoenw@ibr.cs.tu-bs.de>]

>>>>> David Harrington writes:

David> Problematic Requirement: #12 Requirement name: Base Data Types
David> Description of the problem with the requirement: This is a
David> request for clarification of this requirement. The Specs
David> section says that Pointer will probably be needed again. I
David> think Pointer is a new type, since it is not available as a
David> base type, but only as textual conventions in SMI amd SPPI.

I guess there is some confusion here. The SMI has the base type of an
Object Identifier with some derived TCs that turn OIDs into
InstancePointer etc. If SMIng wants to be protocol independent, then
it has to abstract from the protocol differences in instance naming.

In other words, we can not use an OID in the protocol independent
model. Instead, we need to introduce a protocol neutral base type
which we currently call a Pointer. This protocol neutral Pointer is
then mapped into whatever the protocol specific instance naming system
allows us to use.

An example: Consider the SNMPv3 MIBs. They usually use strings to
identify instances. If you want to use COPS-PR to provision these
SNMPv3 configuration tables, you have to introduce Integer instance
identifier. If you need to model a reference between two tables, then
you will end up with an Integer column in COPS-PR while you will end
up with a string valued column in SNMP. The current thinking is to
solve this problem by using an abstract Pointer type in the protocol
neutral module and requiring that the protocol mapping specifies how a
Pointer is mapped.

[Not sure this is understandable. I guess it is getting too late.]

/js