RE: [higgins-dev] Revised Higgins data model goals

["Paul Trevithick" <paul@xxxxxxxxxxxxxxxxx>]

Jim wrote:

> 
> 9 and 10 talk about schema, and seem to say this:
> 
> a) Some schema data model or description language is desired
> b) Schema is tied to a Context, and is applied to everything in that
> Context (I'm inferring this)
> c) Objects (I assume we're talking about DigitalSubjects) have a
> governing schema description
> c.1) An object's schema description governs the data allowed/required
> to exist on the object
> 
> Beyond that, this is not mentioned:
> d) Schema descriptors may apply to data elements of an object
> (attributes/relationships). In other words, schema descriptors can be
> used to govern certain aspects of data. i.e. a SSN may only hold one
> value, a surname may hold multiple values.

That's surely what I meant. I've just added this to [7].

> 
> I'm inclined to envision how some of these goals turn into reality in
> my head, and then argue against the picture I see.
> 
> I agree with (a) above.
> 
> I disagree with (b). It seems unreasonable to tie everything in a
> Context to a fixed set of schema descriptors, especially where a Context
> may be fabricated from the conglomeration of disparate data sources or
> other Contexts.

I agree with you. Sorry, didn't mean to imply what you think I said. It is
clear that we need to be able to support distributed ("composeable" if you
like) schemas. In other words we need the ability to say that within Context
X we use schema s1, s2, s3. And s1 may well include sub-schema s1.1 and
s1.2. I'll add some text about this point.

> 
> I think (c) is fine as long as an object may be governed by an "any"
> schema descriptor. When this happens, an application should be able to
> enumerate and examine each element that makes up the object (rather than
> using a-priori knowledge of what it expects to be there).

Completely agree. I'll take a whack at improving the language
> 
> I assume that the schema governing an object's data elements
> (attributes/relationships), is discoverable given that element's
> identifier. 

Correct.

In other words, an application can see an identifier like
> xyz://foo/bar/surname and know how to access/display its data either by
> some hard-coded knowledge of the "surname" schema descriptor, or by
> looking up the "surname" schema descriptor and discovering things like
> form, sub-elements, allowed meta-data, multiplicity, yada yada.

Yes. This is all possible thanks to goals [2] and [5].

> 
> I also assume that the schema governing an object's data elements
> (attributes/relationships) is independent of the schema governing the
> object which holds it. In other words, a xyz://foo/bar/country element
> behaves the same whether its held on a person object or a device
> object.

Yes. (These are great points that should be included. I'll add them in
tonight).
> 
> Is there a goal in terms of the access of schema? I mean, I see two
> possibilities: 1) Schema is accessed via a Context or the objects held
> with in the Context. 2) Schema is accessed externally (by following the
> identifier's URL, and read using some external protocol like HTTP). I
> think the goals are implying #1

At present the IContext interface requires an implementer to return the
schema in some kind of data stream. I'll add something about this.

> 
> 
> Jim
> _______________________________________________
> higgins-dev mailing list
> higgins-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/higgins-dev