| RE: [higgins-dev] Searching within a context |
|
Jim wrote:
Oh, the URN was just a mock-up of a property name. I assumed properties would be named using (globally) unique identifiers. I didn't assume the property name would include the name of the digital subject of which it's a part.
Got it.
Do we have some example property names floating around?
Well that’s the interesting thing. Our data model doesn’t include any “vertical” ontologies/schemas at all. That is, the kind of ontology that would include a property name/definition for the “eyeColor” that you mentioned. The Higgins ontology defines only the minimal abstract property names required for itself. It does have, for example, “http://www.eclipse.org/higgins/ontologies/2006/higgins#uniqueIdentifier” (which, now that I look at it probably should be more descriptive, like, “#uniqueSubjectIdentifer” or something).
While it is true that every Context Provider developer is free to use whatever schema they’d like, it seems to me that it would be great for there to be a wiki page or something that states the Higgins schema conventions, preferences. It would list a bunch of schemas that developer should use if they don’t have a burning reason to use something different, specific, or even home-made.
With so many standards to choose from, there’s no need to invent our own. Most of the time we just need to take an existing one and wrap it in OWL so that its attribute/property names are expressed as URIs (if it isn’t in this form already, of course), its allowed values are described, whatever other restrictions there are are listed, etc. It would be nice if there was a one para (at most) text description (for humans) too.
For example, if you look at “D3.2: Models” document linked here (http://spwiki.editme.com/Higgins) you will probably find the “eyeColor” you mentioned in one (or more!) of the schemas. If the schema isn’t already in RDFS or OWL then we could just create a wrapper for it.
Jim wrote:
We have the need to be able to perform robust searches within a context. For example, Return all Digital Subjects which have brown eyes, size 10 shoes, but not a dog named fido (unless they also have a cat named felix).
Given our rough consensus on a data model (well two actually), does anyone have a preferred search filter language?
For the high-level API, I can suggest something akin to what LDAP uses because I already know it. The example above would look like this: (&(urn:higgins.property.person.eyeColor=brown)(urn:higgins.property.person.shoeSize=10)(|(!(urn:higgins.property.person.dogName=fido))(urn:higgins.property.person.catName=felix)))
I want understand your "urn:higgins.property.person.eyeColor"-like constructs. Is this one long unique identifier of a property type? Or is this a combination of a Digital Subject reference (e.g. higgins.property.person") and the name of one of that Digital Subject's properties (e.g. "eyeColor"). Assuming it is the former, it seems that the Digital Subject that we're querying is specified somewhere else.
I'm open to anything though, as long as we can filter on property(attribute) values and have sufficient operators (and, or, not, present, equal, ordering, substring, <provider-defined>).
Me too.
Jim |