OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Should xlink:arcrole or xlink:role be the primary way todispatchon related resources? was Re: URIs, names and well known RDDL names,

Thomas B. Passin wrote:
> Jonathan Borden says -
> >
> >     Using Ron's analysis, the arrow or arc connecting the document to
> > related resources is named by the attribute xlink:arcrole. The
> > defines another arrow between the related resource and its type.
> >
> >     To me it seems most natural to have the most important, or primary
> > to name a related resource as an arc from the RDDL document to the
> > resource and that the xlink:arcrole be the name of this arc.
> >
>  This isn't hard to sort out.  Any link on any diagram ***should*** be
> as to its purpose unless there is a clear convention for its use.  This
> rule is often violated, so that people tend to ***think*** that they know
> a link or arrow means when they often don't.  We just have to get clear on
> what the label should be.
> Now, if you draw an arrow representing an xlink from one node to another,
> should be labeled, even if you omit the label  in practice for
> Whatever that label would be,  that's the arcrole.  What could be simpler?
> Jonathan said that the xlink:role could be represented by another arc.
> just going to be confusing.  We don't want two arcs for one xlink, I don't
> think.  If "arcrole" is the label at the middle of the arc, then "role"
can be
> the label for the far end of the arc.

right. this is a better way to describe it.

> According to section 5.5 of the xlink rec, role and arcrole are supposed
to be
> absolute URIs, and the title attribute is there to provide for a
> human-readable description of the "the meaning of a link or resource".
> could be ambiguous - do we title the purpose of the link or the nature of
> resource pointed to?  I say, title the purpose of the link, since that's
> whole reason for having the xlink in the first place.


> This leaves the role.  What should "xlink:role" be used for?  What's left?
> The nature or purpose of the resource pointed to.
> Now for some examples, without syntax.  I hope I've got the right idea
> Let's say that your document is supposed to import two schema fragments,
> a main schema that imports the fragments,  and finally be transformed
using an
> xslt stylesheet.  The stylesheet itself includes another stylesheet.  You
> decide to point to all these resources using an rddl document.

> For the main schema:
> title='Main Schema'
> arcrole='urn:rddl:linkrole:xml-main-schema'
> role='urn:rddl:resourcetype:xsd-schema'                      (or use the
> schema namespace)

Let me also answer Jason's question about why ever have xl:role=xl:arcrole:

In some specific situations, the purpose and nature of a resource are the
same, for example the main XML Schema: Its purpose is to be an "XML Schema"
and it *is* an "XML Schema" so in this particular situation:

xlink:title="XML Schema"

but below the purpose is an "Imported Schema", but the nature is "XML

> For the two schema fragments:
> title = 'Imported Schema'
> arcrole='urn:rddl:linkrole:import-schema'
> role='urn:rddl:resourcetype:xsd-schema'                      (or use the
> schema namespace)


(for us, we do specifically intend to place some documentation of these
terms at the arcrole URI, so we'd actually prefer that its a URL :-)

> This approach covers it all.  The purpose of each xlink is given in human
> machine form,  the nature of the resource itself is indicated by the role
> attribute.  In this example, we can clearly distinguish between the
> and the nature of each resource, even when the same kind of resource is
> for different purposes.

This is nice language.

> Can you see how nicely this would lend itself to an authoring tool?  The
> presents you with a list of known titles.  You pick one, and the rddl
> generator inserts the appropriate arcrole and role attributes (or maybe
> you a choice of allowable resource types, such as RELAX or xml-schema).
> Sweet!

one of the ideas behind basing RDDL on (X)HTML is that an HTML authoring
tool might provide for tag specific extensions.