W3C home > Mailing lists > Public > public-grddl-comments@w3.org > April to June 2007

Re: Comments on GRDDL [OK?]

From: Harry Halpin <hhalpin@ibiblio.org>
Date: Sun, 24 Jun 2007 00:46:03 -0400
Message-ID: <467DF70B.2030303@ibiblio.org>
To: Andrew Eisenberg <andrew.eisenberg@us.ibm.com>
Cc: Dan Connolly <connolly@w3.org>, public-grddl-comments@w3.org, w3c-xsl-query@w3.org, liam@w3.org

(cc'ing Liam the XML Activity Lead, so he knows we are actively trying
to make sure the comment has been addressed.)


    As Chair of the GRDDL WG, we are preparing the specifications and
test-cases for GRDDL to move to PR status, and have noted and discussed
your comment regarding error messages in our telecons.  Jeremy's
original response message is still our official response [1]. Please
read the official response

Since Jeremy's message, we have officially reached consensus on adding
that test-case [2] dealing with 1)  errors  in the transform and  2)
non-RDF results being produced, which we agreed should be distinguished
and dealt with (which is why the test-case has multiple transforms
associated with it).

My interpretation of the test case is (although only the test-case is
normative and should be inspected itself) is that with regards to errors
in the transform and inline with the options you sketched out, we
decided that  we "fail silently" by  not producing RDF results. As
regards non-RDF results being produced, this was decided to be not
allowed by our test-case, i.e. GRDDL should produce only RDF results.
Does this satisfy your comments?

We would like to explicitly know before going to PR.



[2] http://www.w3.org/2001/sw/grddl-wg/td/grddl-tests#error1

Andrew Eisenberg wrote:
> w3c-xsl-query-request@w3.org wrote on 04/30/2007 11:36:28 AM:
> > > 1) When a transformation fails, or it produces non-RDF elements, the
> > > result/meaning should be specified.
> >
> > Without any argument as to why the specification should be
> > elaborated this way, it's difficult for me to take your comment
> > up with the GRDDL WG.
> >
> > We're happy to look at details of any examples or scenarios
> > where this failure would be a threat to interoperability.
> >
> > It seems more likely that the results of our investigation
> > would go into our test cases document than into the specification.
> >  http://www.w3.org/TR/grddl-tests/
> >
> > The specification of a GRDDL result is written declaratively.
> > To say that a transformation fails is like saying that adding 4 and 7
> > fails. Yes, a machine that's doing the computation might
> > fail, but the meaning of the expression 4+7 remains what it is.
> The failure of a transformation could be interpreted by a processor as
> producing no RDF data (fail silently) or the error could be passed
> back to the invoker.
> On producing non-RDF elements, let me extend your analogy and say that
> a processor returns both 11 and "volunteer". The extraneous value
> could be silently ignored, or the entire answer could be considered
> meaningless and the invoker informed of the error.
> We are suggesting that both cases be acknowledged in the
> specification, and that you specify the behavior allowed by a GRDDL
> processor.
>                                                 -- Andrew


Harry Halpin,  University of Edinburgh 
http://www.ibiblio.org/hhalpin 6B522426
Received on Sunday, 24 June 2007 04:46:15 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:52:29 UTC