Re: Test Case Results

> The URL is
>
http://www.w3.org/2000/10/rdf-tests/rdfcore/rdf-ns-prefix-confusion/test0008
.rdf

Dave wrote:
> The file should be renamed as an error case or
> removed entirely
[ ... snip ... ]
> We have discussed 'illegal' rdf: attributes, elements before.  when I
> used similar words to describe what to do when such things were seen,
> the WG did not agree, so they were removed.  So as far as I recall we
> no just considered such things as having no defined meaning.

> At present, since we do not describe what applications do with
> RDF/XML (processing model), we cannot give any requirements on what
> to do with unknown or undefined tokens.

I am happy with the proposal that we limit ourselves to describing legal
RDF, and hence make test 8 an error. It is one of the many inconsistencies
of the old spec that it (mis)specifies the processing of a particular type
of error, but not of other types of error.

Art wrote:
> Based on the argument below, wouldn't rdf:aboutEachPrefix simply
> have no meaning and be ignored?

If we are not specifying processing behaviour of bad input we do not specify
what to do when meeting an attribute with "no meaning". The current version
of ARP, in good faith, based its processing of rdf:aboutEachPrefix on a
particularly reading of para 196 (mine not Dave's!); that processing was not
simply ignoring the attribute. I suspect that trying to specify the
processing of incorrect input is a never-ending task. (It seems hard enough
to specify the processing of correct input and how to differentiate correct
input from incorrect input.) I am not wedded to the current version of ARP,
and only use it as an example, since I understand what it does and why!

Jeremy

Received on Thursday, 6 September 2001 10:16:03 UTC