Re: [Aperture-devel] The future of the Data Source Ontology

[Christiaan F. <chr...@ad...>]

Leo Sauermann wrote:
> That would also require that we know the "version" of the ontology. 
> To mark it, we could: * use a new namespace for the new ontology
> (daring, but this is the right way to do it)

I'd either choose the right way or don't touch it at all :)

> * add a property "thisdatasource.configversiondataformatnumber =
> "1.0" (or something like this) (won't break existing configs)

This sounds like a good idea anyway. Why not use both? The latter
approach tackles the problem of what to do when we want to sligthly
change the namespace sometime in the future.

> Antoni: we don't really have time for the conversion util - how can
> we achieve this?

We might build the conversion util anyway and put it in Aperture.
Doesn't seem to be very complicated (hopefully just map properties from
one namespace to properties in another namespace), only just a nuisance.

>> I would also recommend using namespaces to group the properties per
>>  source, i.e. fileSystemSource:maximumCrawlDepth rather than 
>> fileSystemMaximumDepth.
>> 
> I wouldnt do that, thats too complicated. Rather stick to
> domain/range, thats enough to know. also the ":" in a URI can lead to
> bad bugs...
> having different namespaces for each datasource would require a
> different protege file for each datasource, and a different
> generation script for each Vocab-Constants Java file, which means:
> where today is one file we will have many files for each datasource.
> too  much hassle

But wouldn't it be nice if you would have a single interface holding
all configuration properties of a single source type? I would think it 
would especially make documenting stuff easier. And of course the ":" is 
just a URI using prefix notation, the actual URI would be different and 
not include ":".

I don't really have a good overview of the pros and cons of either
approach, but I'll trust that you know what you're doing.

>> Note that DataSources already have a type attribute, which
>> currently refers to a Class in the Source ontology. I wonder
>> whether this type would become a Class in a particular source
>> namespace, if it would become the namespace itself, or perhaps
>> something else?
>> 
> I don't get this - what do you mean?

Well, that was just me thinking out loud. DataSources (Java objects)
already have a getType method that returns a URI identifying their type
in the RDF world. Right now these URIs model RDFS classes. I wonder
whether this would stay the same or not. When you would introduce a
namespace per source type, you could also use the namespace itself as a
type indicator. Don't know whether this makes sense, just a thought...


Chris
--