Re: Semantics Attributes: XLink CR03062000

At 10:05 PM 11/6/00 +0000, Hartmut Obendorf wrote:
>Hello,
>
>In this second posting I would like to ask a few questions about the
>semantics of XLinks.
>
>What I would like to question is the use of "role" and "title" in the
>spec. As I am not a native speaker, I looked up role in Merriam
>Websters Dictionary:
>
>--
>Main Entry: role
>Variant(s): also rôle /'rOl/
>Function: noun
>Etymology: French rôle, literally, roll, from Old French rolle
>Date: 1606
>1 : [a] : (1) : a character assigned or assumed (2) : a socially
>     expected behavior pattern usually determined by an individual's
>     status in a particular society
>     [b] : a part played by an actor or singer
>2 : a function or part performed especially in a particular operation
>     or process <played a major role in the negotiations>
>3 : an identifier attached to an index term to show functional
>     relationships between terms
>--
>
>I would assume that definition 2 is one that applies to the use of
>role in the spec.

Fair enough; the "role" of a resource would be its type *in the context of 
this link and these arcs*.

>Now, what for what types of elements is role used?
>   a) extended
>   b) simple
>   c) locator
>   d) resource

Also e) arcs, but we were forced to give this one a different name for 
reasons explained below.


>As I mentioned in my previous posting, I don't understand the use of
>role for a link-as-a-whole (a). What is the part and what it the
>process here?
>
>But, more important, I find it hard to understand (c) and (d) as well.
>As I understand it, a link is static. Therefore there is no process in
>the link itself, the only process associated with it is the traversal
>from one resource (or locator) to another resource (or locator).
>
>Since there is no process in the link itself, are the "roles" in the
>spec not rather "types" of resources?

My experience is that "role" is often used as a synonym for "type" in XML 
vocabularies, or at least that when different types-of-types are needed, 
the names "type", "role", and "class" may all be pressed into 
service.  DocBook makes a very precise distinction among these three for 
its purposes, for example.  So I think the use of "role" here is 
legitimate.  However, I can understand how there could be some confusion!

>The only thing I would readily acknowledge as a role is the description
>of the meaning of an arc within the link. Here, the process of traversal
>from one resource (or locator) to another resource (or locator) is
>described. But this is the only attribute _not_ called "role", instead
>the spec calls it "arcrole".

The reason for this is that simple-type elements compress some of the 
functions of extended links into one element, and so in order to keep both 
roles-for-arcs and roles-for-ending-resources, we had to choose a different 
name for one of them.  (XML, of course, doesn't allow two attributes with 
the same name on an element.)

>Once more to overcome my difficulties with verbaliation:
>The link defines an environment in which relationships are described.
>These types of relationships could be called roles. The spec calls not
>only the relationships (given by the arcs) "roles", but also uses "role"
>for explanatory definitions of participating resources.

Hmm, I think so.  I would put it like this:

In order to facilitate sophisticated processing and traversal by 
XLink-aware applications, XLink allows link authors to provide "role" 
information for the following constructs:

- A link as a whole (if expressed as an extended link)
- A starting resource (if it is in an extended link)
- An ending resource (whether in a simple or an extended link)
- An arc (whether in a simple or an extended link)

>The main point given, here is another small question: What is the
>meaning of the "title" attribute for everything but the arc? Along the
>same lines as mentioned above, I don't find a proper answer. Titles would
>almost always carry information about the type of a resource, not about
>its role in the current link (which would be different for another link).

The classic scenario for extended links invented by Tim Bray goes like this:

You are reading a document about the game of Go.  In the text, the word 
"tesuji" (a kind of Go move) is mention.  The word serves as the starting 
resource for several arcs (created in extended links, with this starting 
resource being remote): One takes you to an illustration of the move, 
another takes you to a definition of the word, another takes you to the 
etymology of the word, and so on.

If all these arcs came from random third-party links created by several 
link authors, it might be useful and sufficient to display their link 
titles in a pop-up dialog when the user right-clicks on the word.  But if 
the arcs were specified in the same link, it would not be sufficiently 
helpful to the reader to display the link title; you'd probably want to 
display the titles of the different arcs.

Similarly, two ending resources might (in the absence of arcs) have the 
same title, such as "Person Name", but in the context of arcs, need arc 
titles to distinguish them as "this person's father" and "this person's 
mother".

>If the example in 5.2 "Simple link functionality done with an extended
>link" is correct, the most important title for a link (the only one left
>in simple links) would be the title for the target locator. As I pointed
>out before, I would have thought the main feature describing the
>semantic relationship is the attribute "arcrole".
>
>I would argue that the title of the arc is more important than the
>locator's.

So are you intending to make the case that we should remove the ability to 
specify ending resource roles on simple links?  I believe the WG discussed 
this exact proposal about a year ago, but I'd have to look it up.  If you 
actually make this proposal, that will force me to do so. :-)

>Hope this wasn't too scatterbrained..

Not at all!  I found it helpful.  I hope you did.

         Eve
--
Eve Maler                                          +1 781 442 3190
Sun Microsystems XML Technology Center    eve.maler @ east.sun.com

Received on Wednesday, 8 November 2000 19:09:23 UTC