Re: [xml-dev] RE: Keep business-process-specific data separate?

[Peter Hunsberger <peter.hunsberger@gmail.com>]

On Mon, Feb 2, 2009 at 7:04 PM, Cox, Bruce <Bruce.Cox@uspto.gov> wrote:
> If I understand OED correctly, "animals" is a generalization (generic,
> genre).  A name for a collection of objects with a common property.
>
> Abstract is both a verb and an adjective.  I don't think the verb
> applies here, as someone else suggested.  That I appear to abstract a
> category by inspection of some collection of instances doesn't make the
> category itself abstract, merely derivative.

I think you've hit on a key concept here.  What we're likely doing 80%
of the time when we design Java (or whatever) classes or normalize
schema is likely generalization.  At least on the first couple of
passes.  True abstractions are probably rare. However, I'll note that
a collection of generalized classes can itself form an abstraction for
collecting other data.  Which takes us back to RDF.

For those who might consider this noise, my interest in this isn't
just semantics; rather it ties back to a question Roger implicitly
asked, which is how does one design generalized containers for data.
Experience shows that abstract containers (such as RDF) are hard for
many people to get a handle on.  OTOH, from this discussion I think
I'd vote that  generalizations are somewhat easier to understand. Not
sure that this answers anything, except perhaps why Roger (or the
people he is dealing with) have not jumped on RDF.

[...]

>
> The more generic your schema, the worse the edge cases, I expect.  In
> the case of an abstract schema, I would expect there to be NO edge
> cases, since everything will behave "ideally",

Yeah, probably true; we have a system that uses an almost completely
abstract schema.  You can indeed throw anything at it. However, after
a couple of iterations of descending metadata and paradata many people
seem to wonder if the pain is worth it...

[...]

-- 
Peter Hunsberger