W3C home > Mailing lists > Public > www-rdf-interest@w3.org > January 2001

Re: Klyne Contexts: 3. Statements sets in RDF

From: Graham Klyne <GK@Dial.pipex.com>
Date: Mon, 01 Jan 2001 18:35:30 +0000
Message-Id: <>
To: David Megginson <david@megginson.com>
Cc: "RDF interest group" <www-rdf-interest@w3.org>
At 07:33 AM 12/30/00 -0500, David Megginson wrote:
>Bill de hÓra writes:
>  > Jonas is right but not only for that reason. The RDF numeral
>  > identifier scheme doesn't come with the everyday semantics people
>  > associate with numbers, specifically numeric ordering. They are
>  > just tokens that happen to be confusing to the human reader. Seq
>  > does require ordering though. I like to see this sceheme dropped,
>  > it seems to bend the readability req of XML (it's very readable but
>  > very confusing), but what would replace it?
>There are two alternatives:
>1. revise the spec to declare that RDF statements are ordered to begin
>    with (i.e. a processor that reads an RDF document preserves the
>    relative order of all statements for any given resource); or

Ouch!  I think that is not possible with RDF graph-syntax.  If there is 
ordering, I think it needs to be stated explicitly in the graph.

>2. use resources for property values and add positional information, i.e.
>   <foo:Person rdf:about="urn:xxx:0001">
>    <dc:title>Joe Smith</dc:title>
>    <foo:job-history rdf:parseType="Resource">
>     <foo:job-item dc:title="Intern" foo:pos="1"/>
>     <foo:job-item dc:title="Sales Rep" foo:pos="2"/>
>     <foo:job-item dc:title="Sales Regional Manager" foo:pos="3"/>
>     <foo:job-item dc:title="Booth Babe" foo:pos="4"/>
>     <foo:job-item dc:title="VP Sales" foo:pos="5"/>
>    </foo:job-history>
>   </foo:Person>
>(I wrote the above example in heavily-abbreviated RDF syntax, and will
>leave it as an exercise for the masochistic reader to write it out in
>fully unabbreviated RDF syntax.)

That's a possible approach (though I think it has some of the disadvantages 
of the present mechanism -- how do I add something between "Intern" and 
"Sales Rep"?).

Actually, I think that there are an arbitrary number of ways that ordering 
information can be captured in a graph, and it's not clear to me that any 
one is best for all possible uses.

E.g. in CC/PP, we are proposing a graph structure with "nextXXX" properties 
to indicate the sequence of proxy-derived capability modifications.  This 
effectively, builds the sequence directly in the RDF graph.  I can imagine 
an approach in which 'precedes' and/or 'succeeds' properties are added 
between sequence members.  Real-valued position values are another.  I 
could go on.

Given this wealth of possibilities, all possible using the existing RDF 
base, I don't see why explcit sequencing constructs need to be added to the 
fundamental RDF structure.


Graham Klyne
Received on Monday, 1 January 2001 14:11:10 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:07:33 UTC