Carl, Ron,
I am now confused as to what you guys are talking about. We have a top-down
model, it's the abstract specification. We just don't seem to use it, and when
we do, we tend to abuse it. In the first 5 years, other than writing Simple
Features, all we did was create an information model. There was a glitch when
we went SOA, but that did not affect the information model. GML is built (well
sort of) on that information model. Most of the issues we are having is because
folks have reinterpreted that model, usually badly, or ignored it.
In a way, what OGC should be using as a core is its abstract specification,
especially the volumes that are paralleled in ISO TC 211.
Our major problem is not the lack or a core model, but that we tend to
reinvent or reinterpret our own core model for every document, whether it needs
it or not. And that usually happens after the fact, which is why no one (Ron
and Simon included) seems to understand that everything really is a feature.
The diagram nightmares are the result of shoehorning something into the core
model ex post facto, not a result of its extension.
Regards,
John
John R. Herring
Architect, Spatial Products
Oracle Corporation
One Oracle Drive
Nashua, New Hampshire 03062
ph: 1 603 897 3216
fx: 1 603 897 3334
On Behalf Of Ron Lake
Hi,
I would go further and say that the derivation of such a conceptual architecture -
taking a more top down perspective should be the real work - or at least a very
major component of the OGC work, rather than working only bottom up. I think we
also need to a position in which we make more use of "registration: to add new
features to a specification, rather than by the conventional specification as
document process. This was and remains one of the objectives of GML, and to a minor
extent this achieved through the deployment of a CRS registry - the content of the
registry being (hopefully being blessed by the authority) BEING A standard. I
think this approach can be taken in other areas of our work as well.
WRT a common information model, I think that is a core issue - but not one
easily solved as we have been debating the elements of such a model for a
number of years.
R
