Re: [SE] OOSD note

Bernard Vatant wrote:
> 
> Hello all
> 
> Two remarks on this vey interesting note
> 
> 1. I already made a comment about the definition of OWL classes as "sets" of individuals
> http://lists.w3.org/Archives/Public/public-swbp-wg/2005Oct/0045.html
> No answer so far, but I see it has been included in the OO/OWL comparison table, so ...

I will respond to this later - we'll have another SE phone conference on 
Tuesday...


> 2. [TODO: Possibly also include screenshots of other tools (Cerebra has been suggested) -
> I welcome contributions -- hk]
> Maybe it would be worth considering the new Altova tool Semanticworks
> http://www.altova.com/products_semanticworks.html
> The interesting thing is that this tool has been developed as part of an XML toolkit,
> which makes it quite different of Protégé or SWOOP. The GUI will sound familiar to users
> of XMLSpy, but quite weird to Protégé users. I've downloaded and tried it a bit. There
> seems to remain quite a bunch a bugs in this early version, but worth looking at anyway.

It may indeed be worth to put such a screenshot into the document, even 
if it's only to demonstrate that commercial products are starting to 
emerge.  Tools like SemanticWorks are definitely a good signal (whatever 
their level of maturity), and may provide valuable feedback for the 
future evolution of Protege and other existing tools as well.  Thanks 
for sending the screenshot in the other email.


> To many people in this group, it will be strange to see how it handles "ontology
> validation", which IMO seems to look more like an XML schema validation. "Validation" of
> files edited under Protégé and SWOOP gives strange bunches of "errors", and if you edit
> them, you get of course yet another serialization. I had already experienced that kind of
> problems between Protégé and SWOOP, so it is not big news :))

If you find any file created with SWOOP which Protege cannot load, then 
please let me know (preferrably through the protege-owl list).  As we 
all know there are many different serializations of OWL, including such 
differences like whether to use owl:equivalentClass for defined classes 
or not.  As long as they all mean the same it shouldn't be a problem to 
anyone.


> This leads me to note that maybe there is something misleading in section 3.1
> 
> "RDF just defines the very basic syntax for Semantic Web content, and has an XML
> serialization that allows users to share models on the Web."
> 
> It would be more honest to point that RDF has *many* XML serializations, and that the same
> set of triples can be expressed in an unbound variety of syntaxes, none of them being
> canonical, and which are likely to become arbitrarily complex for large RDF graphs (which
> I stick to think is a major issue for wide adoption and interoperability). So maybe the
> document should include somewhere that tools editors should be aware of issues raised by
> this very variety of syntaxes, and that "conformant" RDF tools (which somehow handle
> internally the semantics of RDF) are bound to exchange RDF not only with each other, but
> with more loosy applications which will rely more on XML structure than on underlying RDF
> semantics, and of which RDF parsers are likely to be less tolerant to exotic
> serializations.
> 
> My concern here is that this group should make the community aware of the risk of building
> software environments able to use only on a specific, and de facto "proprietary" RDF
> serialization. Now that we begin to have a variety of RDF tools coming to the market from
> various backgrounds, it would be good to address real life interoperability issues, like
> "Can I edit an ontology exported from Protégé into SWOOP, Altova Semanticworks ... and
> send it back to Protégé without loosing anything?". Maybe this is not exactly in the scope
> of this note, but seems somehow related.

The same problem occurs between UML tools (which are not really as 
interoperable as they claim to be), simply because each tool focuses on 
a different aspect of the overall language.  In our world, Protege is 
optimized for OWL DL ontologies, and has still trouble with some OWL 
Full or RDF(S) files.  This is because it makes some internal 
assumptions such as "classes and properties are disjoint", which pure 
RDF tools or emacs would not care about.  However, these assumptions 
allow Protege to provide some optimizations in the user interface. 
SemanticWorks does not really seem to care about DL yet, and other tools 
don't have DL expression editors either.

But I generally agree that the freedom and richness of RDF may become an 
issue that will decelerate the adoption of Semantic Web technologies in 
the real world.  The paper should and will not become a plain marketing 
document full of hypes but will mention the risks and current bottlenecks.

Thanks,
Holger

Received on Saturday, 8 October 2005 15:35:41 UTC