[Zope3-dev] Re: RDFLib and Zope 3

[Michel Pelletier]

On Wed, 2005-08-31 at 10:07 -0400, Gary Poster wrote:

> >> I'm interested in contemplating RDF as a full catalog solution for
> >> Zope, at least as a thought experiment.  

Great!

> > Note that the use of bound variables also removes the need for brains.
> 
> We actually don't have catalog brains in Zope 3 anyway, but yes,  
> maybe it would let you only wake up objects when you need to, which  
> is what you are getting at.  It might make sorting easier too.  Need  
> to think this through.

Activation is one issue I actually wasn't getting at, but it's
definitely something to think about.  What I was getting at was that the
result is completely unrelated to any object.  A ZCatalog result object
had the duel role of also being a handle back to the original object (in
addition to preventing activation).  When I work on these things I tend
to disregard the presence of any kind of Python object that the data is
associated with.  I think though that we are well on our way to coming
up with a way to formalize the rdf/object bridge and we'll eventually
have the right solution and terminology here.

> > Sure, there are some notes from me on the z3lab site that might be of
> > interest for thinking about zemantic and catalog integration.
> 
> Could you send a URL?
> 
<snip>
> I think the RDF spec can be used for "interpretation" too, given the  
> rich spellings that RDF allows for predicates, and the typing of its  
> nodes.  I do think that the RDF component should only deal with RDF,  
> so I agree with your general desire.  But RDF is very, very rich: a  
> *lot* of functionality could be in a pure RDF library like RDFlib,  
> including support for predicate constraints and join indexes for  
> instance.  That would be useful whatever your RDF use, if you needed  
> efficiency for common searches.

Ah I didn't get that you were going for such a full solution until this
email.  You're right, it's a bold proposal but I think with some more
fleshing out it can become reality.  This could completely merge the use
cases of both rdf graphs and zope catalogs.

As Dan is interested in componentizing everything, this is a good
oppertunity to think about how to make these changes.

> 
> The extra RDF features could also be a layer on top of a simpler  
> graph implementation, if that were a desired design.  I'm not saying  
> that RDFlib should have all the features, but that a pure-RDF library  
> *could* have all the features.

Yep, I'm on your frequency now.  I think it's a good idea.

-Michel