Re: bug in (approved) test case (syntax)

> On Sun, Jul 26, 2009 at 6:19 PM, Sandro Hawke<sandro@w3.org> wrote:
> >
> >> You are looking at the (advertised) buggy output of the current OWL
> >> API.
> >
> > Yeah, I didn't read the message as nearly that strong. =A0I thought it wa=
> s
> > talking about more subtle issues, with more obscure syntaxes that
> > RDF/XML.
> >
> > Honestly, this makes me deeply skeptical about ending CR.

To be clear, I'd vote "-0" not "-1".  Hopefully more data will move me
to "+1" in the next few days.

> Why -  want to unpack this. Is it because the OWLAPI that we are using
> on the test site should be an exemplary implementation?

I suppose it's the combination of (1) my suspicion that OWLAPI is the
only implementation of mapping-to-rdf in use, and (2) realizing that
OWLAPI has bugs.  (Plus a bit of (3) my own anxiety, as I implement the
mapping, of how error-prone a bit of code it is.)  Hopefully someone
will step forward with confirmation of other independent, serious,
well-tested implementations of the mapping.

> There is an OWLAPI v3 that is, as I understand it, almost ready.
> Perhaps it could be swapped in to the manchester site we are using for
> these translations.

It looks to me like the dynamic service (which has the bug) is using v3.
It's definitely not using the trunk version of v2 (owl1.1), since that
OWLXML is quite old (wrong namespace, wrong attributes, etc).

> > But are they right? =A0 Why would I think they were?
> 
> Because, as the page says, they are manually translated to RDF and are
> normative,

Were they really manually translated?  I figured the wiki would store
and display whatever was done by hand.  Who did the translation, and
where is the master copy stored?  (I'd expect manual translation to be
error prone as well.)

> and if they have been approved there have been two
> implementations that have passed the test, which was our decided upon
> criteria.

But in this case, the two implementations both used OWLAPI for the
parsing -- I'm pretty sure Mike Smith talked about running them through
an OWLAPI test harness.

> I do think the web site should be cleaned up - we shouldn't offer to
> dynamically translate to RDF if we have a hand written version, and we
> should perhaps just remove the links to the service until it works
> adequately.

Yeah, or label it more like: "Experimental service with significant
known bugs.  Please see _bug list_ and report more bugs _here_."

    -- Sandro

> >>
> >> On Sun, Jul 26, 2009 at 1:49 PM, Sandro Hawke<sandro@w3.org> wrote:
> >> > Looking at a random test case:
> >> > =3DA0 =3DA0http://km.aifb.uni-karlsruhe.de/projects/owltests/index.php=
> /New-Fe=3D
> >> ature-ObjectQCR-001
> >> >
> >> > I find the provided RDF/XML (attached) isn't well-formed XML. =3DA0I e=
> xpect
> >> > the problem is pervasive.
> >> >
> >> > FWIW, the version in the all.rdf, which is somewhat different, looks
> >> > okay. =3DA0(It's also attached.) =3DA0I suppose we're running some ble=
> eding e=3D
> >> dge
> >> > converter to produce the RDF/XML?
> >> >
> >> > =3DA0 =3DA0-- Sandro
> >> >
> >> >
> >> >
> >

Received on Monday, 27 July 2009 13:37:59 UTC