Re: Fresnel: Lenses: summary and unsolved issues

From: Emmanuel Pietriga <Emmanuel.Pietriga_at_lri.fr>
Date: Mon, 25 Apr 2005 17:15:15 +0200

Jacco van Ossenbruggen wrote:
> Emmanuel Pietriga wrote:
>>
>> 1. Jacco's wondering about the possibility to define lenses as
>> refinments of other lenses. We do have such a mechanism. See section
>> 5.2 Lens Inheritance [2]. Does this meet your needs?
>>
> Yes, this is what I meant. I missed section 5.2, sorry. Technical
> question: why not use RDFS built-in semantics for inheritance using
> subclass and subproperty hierarchies?
> I assume you will have good reasons for it - just wondering.

One possible answer to your question (but I'm not sure I understand what
you mean exactly) is that RDFS subclass/subproperty relations are used
to describe relationships between property types and classes in the
context of the description of an RDF vocabulary. So they act on
schema-level concepts, not data-level instances.

So I do not really see how these could apply here. What we need is a way
to relate lens instances. We do not have different classes of lenses.
Just lens instances that are all of rdf:type fresnel:Lens.

Maybe an example of what you propose would help.




>> 2. How do we resolve conflicts between different lenses that apply to
>> the same instance? This is indeed still underspecified, and we need to
>> work on this.
>
>
> Related thought (Lloyd deserves the credits): you could view fresnel
> lens and style info as "just" another type of metadata, which you can
> attach to any web resource like other RDF. Note that currenty, the lens
> specifies to what instances (say, instances of foaf:Person) it applies
> by attaching a fresnel:lensDomain property to the lens. You could also
> imagine you attach a fresnel:lens property to the foaf:Person class to
> specify a suitable lens. This would also allow use of rdfs
> subclass/subproperty inheritance: instances of subclasses of foaf:Person
> would inherit the lens, and a subproperty of fresnel:lens, say,
> fresnel:defaultLens, would return a valid result even when foaf:Person
> is queried only on fresnel:lens.

Interesting idea. Instead of saying "this lens applies to this", you say
"to display this you can use this lens". One problem with this is that
you cannot say "I want this lens to apply to all instances of
foaf:Person that have a ex:age of 30". But lensDomain and this method
could coexist, as rules and style attribs coexist in HTML+CSS.



> Where the current approach is similar to a CSS style rule that uses a
> selector to specify to which HTML classes it applies, the other approach
> would be similar to, for example, an HTML element that defines its own
> style. But thanks to the properties of RDF, without many of the
> drawbacks of the HTML case!
>
> In HTML, this approach is a maintance problem because it becomes hard to
> change the style when it is defined inline in the HTML source.

That's why one should refrain from using style="".


> But in
> RDF, the direction of the property does not have any relation with the
> files in which the property is specified. For example, the triple that
> relates the foaf:Person class with the fresnel lens does not need to be
> in the same file as the definition of foaf:Person or the lens.

> A second drawback of the inline HTML approach is that you need to
> duplicate the same style specification into every element that uses it.
> But in RDFS, you can make use of the instance/class relation (a style
> defined on the class level applies to all instances) and of the subclass
> hierarchy (a style defined on the class is inherited by the subclasses).

Yes indeed. So I agree that this proposal does not bear some of the CSS
style attrib drawbacks. But considering (something you may have missed
in the manual) that FSL and basic Fresnel selectors are RDFS-aware (and
will thus select instances declared as belonging to subclasses of the
class declared by a selector), what is the advantage of expressing lens
domains this way? I'm not saying there is none, I just do not see it yet.

-- 
Emmanuel Pietriga
INRIA Futurs - Projet In Situ    tel : +33 1 69 15 34 66
Bat 490, Université Paris-Sud    fax : +33 1 69 15 65 86
91405 ORSAY Cedex            http://www.lri.fr/~pietriga
Received on Mon Apr 25 2005 - 15:16:23 EDT

This archive was generated by hypermail 2.3.0 : Thu Aug 09 2012 - 16:39:18 EDT