RE: NIM requirements/conventions for textual representations

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

Hi Juergen,

> >>>>> Durham, David writes:
> 
> David> I am wondering if the textual representations requirement
> David> should be left as stated. Most of the people I've talked to
> David> consider UML (or a subset thereof) a relatively agnostic way to
> David> represent the model. The textual representation of that might
> David> as well be MIBs, MOFs, pure ASN.1, XML, etc. So perhaps, the
> David> textual representation requirement should be rephrased to imply
> David> a textual mapping of the UML. The real requirement then is that
> David> there should be algorithmic mappings standardized for two or
> David> more of the most-loved data representations. The NIM working
> David> group, then, would be responsible for providing the standard
> David> mappings documents.
> 
> First, I think the requirements document should not already make any
> decisions which technology to use.

[Dave] And they don't. However, the requirements currently specify that
there MUST be a textual representation for the model. This is the
requirement I am double-checking with the list since not all representations
need be textual.

> 
> Second, what do you mean when you talk about UML? 

[Dave] The subset (or perhaps superset:) ) of UML that meets the NIM
requirements. 

> UML is a pretty
> complex beast with many diagram types and things like the object
> constraint language. 

[Dave] I will also note that constraints are currently part of the
requirements. So OCL is certainly not out of the question.

> Even if many people agree that UML is a good
> thing, they may have completely different opinions about what UML
> actually means to them.
> 

[Dave] Agreed. The question should be, what UML means to NIM. Again, the
requirements do not take a stand on what should ultimately be used to
represent the model. That is clearly for a NIM WG to decide once the
requirements are agreed.