Putting final touches on role interpretation for RDF clarity

[Forwarding this to get it into the comments archive, as I use it for my 
editing "marching orders"...   -elm]

>Date: Fri, 18 Feb 2000 12:41:27 -0500 (EST)
>From: Dan Brickley <danbri@w3.org>
>To: "Eve L. Maler" <elm@East.Sun.COM>
>cc: w3c-xml-linking-ig@w3.org, connolly@w3.org, swick@w3.org
>Subject: Re: ACTION: DV to arrange a careful review of RDF mapping(s)
>
>
>[copying Ralph and DanC; thread
>context: 
>http://lists.w3.org/Archives/Member/w3c-xml-linking-ig/2000Feb/0009.html ]
>
>On Fri, 18 Feb 2000, Eve L. Maler wrote:
>
> > I think I see the problem now.  I'm not going to quote your message,
> > because I want to get right to the point: We've been fuzzy on the exact
> > role of role, so to speak, but I think we can fix it.  How about this?
>
>Yes, I think this is the heart of the problem.
>
> > A role on a resource is a description of that resource *as an ending
> > resource* in that link, with the understanding that this is not the
> > be-all-and-end-all description of it.  It might have different guises in
> > other links, or when used for other purposes.
>
>I like this. It doesn't rule out applications exploiting additional
>conventions allowing firmer conclusions to be drawn when encountering
>these, but it does warn that in the general case the contents of role on a
>location-type resource will be unpredicatable without additional info...
>
>eg.  <loc xlink:type="locator" xlink:href="Smith" xlink:role="mytypes:male" />
>
>So... I _might_ make it known (by my own means) that the presence
>of values from the 'mytypes' namespace implies an RDF 'type' arc from the
>targeted resource to the resource whose URI reference is gotten by
>concatenating the URI of the 'mytypes' namespace with the qualified name.
>In other words, custom apps could dig for more RDF-esque data if they had
>advance knowledge of the role values, but no general RDF interpretation
>strategy is provided by Xlink itself for roles on locators. Which seems
>right to me.
>
>
> > A role on an arc is a description of that arc's *ending resource* relative
> > to the *starting resource*.
>
>This also feels right. I'd try a slight variant to stress that it's the
>inter-relationship that's key, not some property of the ending resource:
>
>  "A role on an arc describes a relationship that holds between an arc's
>*ending resource* relative to the *starting resource*.
>
>Once this is set up, I think the RDF mapping could be done in a couple of
>sentences, something along lines of:
>
>"This corresponds to RDF's notion of a property, and more
>generally provides for directed labelled graph models of XLink arc
>roles. When modeled as directed graph, XLink arcs are labelled with the 
>role name, and
>can be interpreted as stating that "[source-resource] HAS [arc-role]
>[target-resource]".
>
>(Conversely: " [arc-role] OF [source-resource] IS [target-resource].")
>
>Some examples:
>
>(do the  "_ HAS _ _ ", and "THE _ OF _ IS _" interpretation seems plausible?):
>
>         [source.xml] HAS sitemap:helpDocument [target.xml]
>         [WGHomePage] HAS W3UTIL:issueList [WGOpenIssues]
>         [WGHomePage] HAS W3UTIL:mailListIndex [http://lists.w3.org/.../etc/]
>         [SomeDoc] HAS crit:onlineCritique [SomeCriticalDoc]
>
>
> >       For example, if you say:
> >
> > <loc xlink:type="locator" xlink:href="Smith" xlink:role="myco:male" />
> > <loc xlink:type="locator" xlink:href="Jones" xlink:role="myco:female" />
> > <go xlink:type="arc" from="myco:male" to="myco:female" 
> role="myco:daughter" />
> >
> > ...then the "to" resource is the corresponding "from" resource's daughter.
>
>Yep
>
> > Does this help?
>
>I think so. XLink looks pretty handy; I can make all sorts of (possibly
>libellous) RDF assertions in a .sig file's worth of space...
>
>         I realize that because the arc-type element
> > "batch-creates" arcs, it may be painting several actual arcs with a broad
> > brush.  However, this can always be remedied with suitable application of
> > lots of different role values.
>
>This can also be quite a useful facility, eg. for sitemap-type apps
>
>Dan
>
>--
>danbri@w3.org

--
Eve Maler            Sun Microsystems
elm @ east.sun.com    +1 781 442 3190

Received on Friday, 18 February 2000 13:00:33 UTC