W3C home > Mailing lists > Public > www-rdf-comments@w3.org > October to December 2001

Re: aboutEach

From: Brian McBride <bwm@hplb.hpl.hp.com>
Date: Tue, 20 Nov 2001 09:51:51 +0000
Message-Id: <5.1.0.14.0.20011120093305.02bafad0@0-mail-1.hpl.hp.com>
To: Roland Schwaenzl <Roland.Schwaenzl@mathematik.Uni-Osnabrueck.DE>, www-rdf-comments@w3.org
Cc: roland@scarlett.mathematik.Uni-Osnabrueck.DE
Hi Rolan,

Thanks for the comments.

At 23:38 19/11/2001 +0100, Roland Schwaenzl wrote:


>1. CARA has implemented "aboutEach".


Cool.  How have you implemented it?  Does it give you any scalability 
problems?  Given a large source stream with containers, will you eventually 
run out of resources, e.g. memory. If so, do you consider that to be important?


>2. I don't think rdf:aboutEach requires processing of sub-property relations:
>
>    Members of containers must have rdf:_nnn arcs by RDF M&S -
>    without any inferences required to deduce something like that (see RDF 
> M&S End of sec5).

I think the issue here is what if some declares ex:foo to be a subproperty 
of rdf:_1.  How would you expect the following to be processed:

   <rdf:Bag rdf:ID="bag">
     <ex:foo rdf:resource="http://ex"/>
   </rdf:Bag>
   <rdf:Description rdf:aboutEach="#bag">
     <ex:bar>bar</ex:bar>
   </rdf:Description>

It would be possible to define aboutEach to only work on rdf:_nnn 
properties and not their subproperties.  Is that what you had in mind?


>3. Competing use of explicit numbering (rdf:_1 ...) and rdf:li - as 
>"Refactoring"
>    seems to allow - is awkward, confusing and yields erronous results:
>
>    <rdf:Bag rdf:_1="one">
>    <rdf:li>one</rdf:li>
>    </rdf:Bag>
>
>    ARP constructs from it 3 triples: Get [genID --rdf:_1--> one] twice. 
> Statements in a model form a set, which has triples as elements.

True.  Does Cara eliminate all duplicate statements?  For example:

   <rdf:Class rdf:ID="Foo"/>
    ...
    100M other statements
    ...
   <rdf:Class rdf:ID="Foo"/>

As for erroneous results; yes you can write bad RDF.  There is a tradeoff 
here.  The grammar was clearly was ambiguous.  It was felt that on balance, 
this way of cleaning it up was best.

>    The construction currently is not equivalent with
>    <rdf:Bag>
>    <rdf:li>one</rdf:li>
>    <rdf:li>one</rdf:li>
>    </rdf:Bag>

That's right they are not the same thing.


>    which get's correctly processed as [genID --rdf:_1--> one, genID 
> --rdf:_2--> one].
>
>    Explicit numbering is of fairly limited use. Is there any format 
> specified for explicit numbering?
>    ARP excepts rdf:_-1 without any complaint.

Thanks for that one.  I'll pass it on to Jeremy.  Whether its legal or not 
depends on the outcome of another issue we haven't resolved yet; what to do 
about names in the rdf namespace the parser does not recognise.  But I'm 
sure Jeremy would want to output at least a warning.


>4. Which thoughts the wg has about xml:lang ?
>    It should either be removed from RDF or should create triples -


Do say why, maybe in a separate thread.  The WG is trying to get its 
collective head around the nature of literals.  We have looked at M&S and I 
think that it's is pretty clear that the language is 'part of' a literal, 
according to M&S.  The community seems to have a lot of trouble accepting 
that.  If you have a killer example/argument either way, that would be 
really helpful.

Brian
Received on Tuesday, 20 November 2001 04:52:02 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:15:17 UTC