RE: assymetric reference of properties

Seems like there might be another solution. This solution begins to step
away from "orthodox" RDF and points to taking first-class properties
seriously. It involves making the property an object in its own right. In
other words, rather than saying "x is identified by y" and "y identifies x",
one would say something along the lines of
"instance-of-identification-property connects x and y". 

Jeff

-----Original Message-----
From: Eric Hellman [mailto:eric@openly.com]
Sent: Thursday, January 27, 2000 7:46 PM
To: martin; www-rdf-interest@w3.org
Subject: Re: assymetric reference of properties


We encountered the same situation in our linking metadata schema. A 
lot of code was expended making sure that we don't get stuck in 
endless loops while traversing models with these reciprocal 
relationships- probably its a good excuse to keep the "inverse" 
property out of rdfs. But I certainly agree that it's a needed 
facility at some level.

Eric

At 11:48 AM +0200 1/26/00, martin wrote:
>Dear collegues,

<stuff deleted>

>In the CRM we have foreseen an "inverse name" for all data models, which
>do have the above problem. I.e. as convenient, users may either use
>the name "is identified by" for the property in the first example,
>or the name "identifies" for the its inverse. The simultaneous provision
>of both links as above, without further formal semantics, seems not
>to be very satisfactory.
>
>I can imagine 3 solutions:
>
>a) A statement is introduced in RDFS, stating that property B is the
>    inverse of property A. This would allow at least to formally exchange
>    information about the inverse equivalence of A and B.
>b) RDF descriptions are extended to declare property instances of the
>    inverse of a property type. This would avoid double declaration,
>    lead to redundance-free models, but make an arbitrary decision to
>    from where the "original" property is read. The use of the inverse
>    property type becomes less comprehensive to read for humans.
>c) RDFS is extended to register two names for each property, a forward
>    and a backward one, as we propose in the CRM, and RDF descriptions
>    allow for either use according to the direction of use. The latter
>    seems to us the most appropriate solution.
>
>Comments welcome.
>
>Martin Doerr

Eric Hellman
Openly Informatics, Inc.
http://www.openly.com/           21st Century Information Infrastructure
LinkBaton: Your Shortcuts to Information  http://linkbaton.com/

Received on Friday, 28 January 2000 12:01:59 UTC