Re: [glue-wg] [ogsa-wg] Nov 29 - Info Model call minutes
Donal K. Fellows ha scritto:
Sergio Andreozzi wrote:
Ellen Stokes ha scritto:
One observation on your modified diagram....what do you mean by 'Resource Model'?
In OGSA, we use the terms information model (abstract) and data model (concrete)....we do not use or define resource model since it is an ambiguous term.
yes, Information Model is a better term in that case. GLUE is a possible instance. I attach the modified version.
I think you need to be careful with GLUE since that's very close to defining both an information model and a data model, making it easy to get confused. For example of how they differ, an information model might say that a particular class supports an extensible list of properties (other than the standard set it defines) and the data model would say what format that list of properties must take (e.g. XML document, Java properties format, comma-separated list, etc.) It's the information model that it is important to standardize at the architectural level, since it is fairly easy to code up data model transformations (though those matter much more for specific protocols). GLUE is an information model of Grid resources. This is the main goal and nature.
The aspect of extensibility is secondary to this and is present because is a requirement from many parties. The fact that the abstract GLUE Information Model (represented using the object-oriented constructs of UML Class diagrams) is rendered on different concrete data models (XML Schema, relational and LDAP) makes it difficult to delegate at the concrete data model level the extensibility aspects. "Hooks" are currently present at the information model level and can be used to advertise information that is not currently captured by the specification without waiting for a new spec version. The GLUE Spec is deliberately not rigorous in defining the format of the content of the extensibility points. Cheers, Sergio -- Sergio Andreozzi INFN-CNAF, Tel: +39 051 609 2860 Viale Berti Pichat, 6/2 Fax: +39 051 609 2746 40126 Bologna (Italy) Web: http://www.cnaf.infn.it/~andreozzi
Sergio Andreozzi wrote:
The aspect of extensibility is secondary to this and is present because is a requirement from many parties. The fact that the abstract GLUE Information Model (represented using the object-oriented constructs of UML Class diagrams) is rendered on different concrete data models (XML Schema, relational and LDAP) makes it difficult to delegate at the concrete data model level the extensibility aspects. "Hooks" are currently present at the information model level and can be used to advertise information that is not currently captured by the specification without waiting for a new spec version. The GLUE Spec is deliberately not rigorous in defining the format of the content of the extensibility points.
My point was that the information model should state what (in general) the information is (e.g. a set of properties) without stating the format of that information (which is the data model's job). At least back in the GLUE session in OGF21 those two aspects were a bit mixed up. (Or at least as far as I could see at the time; I may have got a false impression. :-)) This isn't to say that the data model shouldn't be worked on. Just that the split between the two is important. Donal.
participants (2)
-
Donal K. Fellows -
Sergio Andreozzi