Re: Re 2: [LC Response] to Jan Wielemaker, Re: Triples and OWL2

Dear Ivan, working group,

[ This reply is a bit messy because I accidentally sent my initial reply
  to Ivan personally instead of using the list.  Ivan replied to me
  personally. I left this discussion in this mail for reference and summarise
  my findings below.
]

My first complaint was the unclear role of OLW/XML vs. RDF/XML. Ivan
acknowledged this and promised improvement. He made this more concrete
in the private exchange (see below). After reading the
Document_Overview, the role of the various languages is now sufficiently
clear to me.  So, I'm satisfied about the poposed changes to the
presentation.

My second point was that OWL/XML causes confusion (two XML
serializations for practically the same thing) and introducing a second
OWL/XML syntax harms the possibility to use `cherry-picking' of OWL(2)
features into RDFS applications. I think this path from shallow
semantics, slowly incorporating parts of OWL towards real OWL
applications is crucial to the success of the semantic web. 

Ivan understands this, but claims that OWL/XML is needed to get to an
OWL syntax that conforms to XML toolchains (such as XSLT, editors,
etc.).  Otherwise OWL/XML has the same status as e.g., Turtle.

I can understand the problem of XML tools, but I think this must be
solved by the/an RDF/XML WG, and *not* by the OWL WG. If we take a look
at the diagram in http://www.w3.org/2007/OWL/wiki/Document_Overview, we
see 3 OWL syntaxes and 2 RDF syntaxes! In addition we see 2 models:
OWL2/Structure and OWL2/RDF. As a tool-developer, this is getting
`babylonial proportions' ... Soon we will see applications like
ImageMagic (bitmaps) or gpsbabel (GPS data) for the semantic web.

I'm totally happy to whatever and syntax and model tools and comunities
want to use internally, but why must all these alternatives be described
as part of the standard? It is ok to use the abstract/functional syntax
to simplify describing the semantics of OWL, but please do not propose
it as a concrete syntax.  It can (and does, see for example Thea,
http://www.semanticweb.gr/TheaOWLLib/) play a role in defining an OWL
API for a programming language. This is totally fine.

I fear that presenting the SW using so many practically isomorphic
languages harms the initiative.

 Regards --- Jan


On Thursday 26 February 2009 11:04:13 pm you wrote:
> Hi Jan,
>
> let me give me my private perspective as a response; this may not be
> exactly the Working Group's response and it is certainly not official,
> but it may still help for a better understanding...
>
> Jan Wielemaker wrote:
> > Dear Ivan,
> >
> > Thanks for the reply. It is a relief to understand that the formal role
> > of RDF/XML with regard to OWL has not changed. That said though, the
> > current presentation points in a different direction and in my opinion
> > the required changes are so large that I cannot say that I am satisfied
> > without reading the next release of the documents.
>
> I can understand that. Let me, nevertheless, give you some pointers
> based on the official results of the f2f meeting we had earlier this
> week. Maybe that will give you some confidence:-)
>
> - It has been decided to add a 'Document Overview' document to the whole
> set. This gives a high level overview of all items, including the role
> of OWL/XML. The editor's draft is at:
>
> http://www.w3.org/2007/OWL/wiki/Document_Overview
>
> Although not directly relevant to this issue, there is also a roadmap at
> the end of that document that will be repeated in all documents
>
> - The structural specification currently has a bunch of examples but all
> in its own functional syntax. In the new version all examples will be in
> this functional syntax _and_ in RDF. Well, not RDF/XML but probably in
> Turtle, but that still reinforces the role of RDF. That is not yet
> visible on the Web, the editor has to do it.
>
> - I guess some extra text should be added to the intro part of the
> OWL/XML document to reinforce the issue...
>
> > Some details on your reply:
> >> The confusion obviously comes from the fact that the OWL/XML syntax,
> >> which was published as a note[1] for OWL 1, is now on Recommendation
> >> track. OWL/XML for OWL 1 was an optional feature that OWL 1 tools could
> >> implement if they wished to do so. The fact that OWL/XML is now planned
> >> as a recommendation has not changed this.
> >
> > No, but it opens a different perspective. I'm not totally fluent in the
> > W3C terminology, but at least it gives the impression that the OWL/XML
> > might have an official status in the future.
> >
> > I fear that having OWL/XML next to RDF/XML will cause serious confusion
> > and give few benefits in return. I think it is confusing and
> > inappropriate for this working group to plug OWL/XML.
> >
> > My position is based on two observations. First of all, two similar
> > languages to serve the same purpose is confusing and aks for extra
> > implementation effort. Although RDF/XML allows for ugly XML, many
> > constructs can also be expressed in a reasonable way. If OWL2 requires
> > constructs for which this does not apply, this should be negotiated with
> > RDF/XML. Second, many users of the Semantic Web use RDFS and do
> > cherry-picking of OWL(2) features. The RDF/XML serialization makes this
> > trivial. A secondary OWL/XML will make this much more complicated. This
> > possibility for cherry-picking however creates a smooth path from
> > shallow semantics in meta-data and converted databases to more
> > challenging semantics in OWL(2). In my experience, maintaining a
> > continuous and smooth path is vital to the success of the Semantic Web.
>
> I do understand what you say, but I may not agree with everything...
> Just let me add some issues to your consideration
>
> - the fact of having several serializations around is, by itself, not an
> issue that creates too much difficulties. Many people use turtle and I
> regularly get requests to standardize turtle. It is a de-facto standard
> in the RDF world and everybody is fine with it.
>
> - to avoid any misunderstandings: there is _nothing_ in OWL 2 that could
> be expressed in OWL/XML but could not be expressed in RDF (ie, turtle or
> RDF/XML). As I said in my original mail, there might be some corner
> cases of RDF graphs (that are perfectly ok RDF graphs!) that cannot be
> put into RDF/XML but, luckily, turtle can come to rescue. Ie, OWL/XML
> does not represent any extra expressivity whatsoever. I hope the new
> Document Overview makes this fact clearer.
>
> - tools, implementations, databases, etc, are _not_ required to
> implement OWL/XML. This also means that nobody forces the 'cherry
> picking', as you call it, to be done with OWL/XML. Put it another way:
> OWL Full users can safely ignore OWL/XML, so can implementations and
> toolsets.
>
> - so why OWL/XML, you might ask? Well, the issue is that there are
> ontology applications that are closely bound to the XML infrastructure,
> mainly in big entreprises. These may want to rely on xml based
> workflows, XML schema based editors, etc. It may be a somewhat niche
> market but, for those, RDF/XML really breaks down for well known reasons
> (doing an XSLT script for RDF/XML ain't pretty...). Ie, having an XML
> serialization of OWL ontologies that could be used in those
> infrastructures is important. To have it fixed and interchangeable is
> also important, and that is what OWL/XML is made for. And for nothing
> else, in fact...
>
> I hope this helps to dissipate your doubts at least a little bit...
>
> Sincerely
>
> Ivan
>
> >  Regards --- Jan

Received on Monday, 2 March 2009 10:01:49 UTC