W3C home > Mailing lists > Public > w3c-sgml-wg@w3.org > May 1997

Re: Comments on XML-Link Spec

From: Peter Murray-Rust <Peter@ursus.demon.co.uk>
Date: Tue, 20 May 1997 17:28:36 GMT
Message-Id: <6930@ursus.demon.co.uk>
To: w3c-sgml-wg@w3.org
In message <199705171509.LAA05066@ids2.idsonline.com> "Eric Johnson" writes:
> My comments below may more prove my ignorance than be enlightening.  So be

I am in the same position as Eric - I have never used HyTime or TEI and 
my knowledge is completely confined to the discussions on this group.
I suspect that I have severely misinterpreted parts of the spec.
I also find it difficult to understand abstractions without concrete 
implementations to help me, which is why I think (a) implementations and
(b) examples are critical at this stage.  I believe that the current spec is
not easy to implement and that unless some of the points that Eric has 
described are addressed, the July 1 version will not be crisp enough to
get multiple implementations that have similar behaviour.

> it.  I'm working from the white printed version of the spec Jon handed out
> at the WWW6 conference.  With these comments, I'm coming from two
> directions: greater clarity where possible and spec independence.  I
> applaud the brevity.  However, the spec, it seems to me, as much as
> possible must stand on its own and minimally rely on references to other
> specs.

Agreed.

> 
> 1.4 TERMINOLOGY
> 
> It becomes clear, but not immediately, for example, that a link can have
> more
> than one resource.  I kept having to refer back to the definitions, which
> seem sometimes incomplete, in trying to understand their full implications.

As far as I can see - and I've posted on this - a link can can 0-N resources.
The number of resources is not determinable from the locator because it
may depend on the result of a TEI Xptr search.  If a resource is a Set of
elements/sub-elements etc., then it should be called a Set.

> 
> RESOURCE-  In the case of 'resource,' can a link have more than two
> resources?  Don't see anything in the spec that suggests a link cannot.  If
> so, is there an upper limit?  Logic suggests only two, but perhaps this
> isn't overtly logical.  Does a link have to have a resource?  The spec
> seems to imply that it does and logic says it must.  Is a point therefore a
> resource?  E.g., a location to insert a pointer or a resource?  

There is a potential confusion here because a link may be SIMPLE/LOCATOR
which locate a ResourceSet, or it may be EXTENDED when it contains a 
LocatorSet.  Each member of the LocatorSet locates a ResourceSet.

> 
> LOCATOR-  "may be used to locate a resource"  If a link involves more than
> one resource, is the 'a' perhaps a source of confusion?  Could there be
> clearer wording?  What other purpose specifically could a locator serve
> than to locate a resource?  An example or two?

I think examples are essential.  Then we can comment on what we think they
(a) mean (b) do - if anything.  Also diagrams help a lot.

> 
> LABEL-  Here defined as a "caption", later labeled as TITLE.  If a caption
> is "associated with a resource," cannot a link have more than one label? 
> If not, why not?  And please make the possibilities clearer.

I assume that each Locator can only have one TITLE (attributes cannot be
duplicated).  

> 
> TRAVERSAL-  I'm hung up on the 'a' in "accessing a resource."  Where does a
> traversal begin?  The spec seems to say where under the next item,
> MULTI-DIRECTIONAL LINK, so why not here?

I think there is confusion in the use of the wors LINK.  An EXTENDED
element is not a link per se, but a LINK-CONTAINER.  It can contain 0-N
LOCATORs.  In the special cases:
	0 represents a null link
	1 is essentially a simple link (its 'anchor' being at the position
		of the tree/stream within the EXTENDED element (otherwise 
		where is it positioned?
	2 represents two links.  These might either form a 'double arrow'
		or be two independent links starting from the position
		in 'EXTENDED'.  If the implication is that there is a 
		reciprocity, it's not spelt out.
	3 can only reasonably be a trident-shaped link?
	and N likewise.

In fact, as I write this, I'm getting concerned that JUMBO doesn't do what 
other people expect it to do.  Please comment on the following:

<P ID="one">
This is para one
</P>
<P ID="two">
and this is para two.
</P>
<LINKSET XML-LINK="EXTENDED">
This linkset has two ends
<LINK XML-LINK="LOCATOR" HREF="#one">
<LINK XML-LINK="LOCATOR" HREF="#two">
</LINKSET>

Now - I assume that the intention of the authors is that if I'm sitting on
"one" and click it it jumps to "two" and if I'm sitting on "two" it jumps t
one.  That's not what JUMBO does, and I suspect it's wrong.  [JUMBO treats
the links as a multi (bi-) headed arrow.  If XML-LINK was ACTUATE="AUTO"
it would immediately traverse to both and (say) light them up.

If JUMBO is wrong, and I suspect it is, then I would find it difficult to
see how there could be less than two LOCATORs in LINKSET.  I can see how there
could be more than 2 - this represents a complete graph for N components, 
although I don't see that traversal is a meaningful activity here.

> 
> MULTI-DIRECTIONAL LINK-  The wording here clearly implies to me that a link
> may have more than two resources.  If that's what's intended, that's what
> comes across.  If true, please so state.
> 
> IN-LINE LINK-  The examples are what make the definition clear.  Perhaps a
> phrase to explain why it can serve as its own resource?

'A link which serves as one of its own resources'. 
[resource] 'anything which happens to be reachable by the use of a locator
in some linking element'.
This looks wrong:

<A ID="foo" HREF="#foo">

would qualify, but is presumably not intended.

> 
> OUT-OF-LINE LINK-  Does "multi-directional" here truly mean
> multi-directional, or is it only bi-directional?

I suspect that bidirectional will be the commonest.  Multi-bidirectional
could be composed of lots of bidirectional links.

> 
> 1.6 TYPES OF LINK TYPES
> 
> LINK TYPOLOGY-  "varying numbers of resources" seems vague.  More precision
> possible without hampering the implementer of the standard?
> 
> FORMATTING-  In what standard, if any, will link formatting information be
> contained?

This will be application-dependent, I suspect.

>

HenryT has mentioned an 'implicit resource'.  Henry, could you elaborate 
because I don't understand where this is required.

> 

Independently, I am concerned about when Traversal takes place and may post
later.

	P.

-- 
Peter Murray-Rust, domestic net connection
Virtual School of Molecular Sciences
http://www.vsms.nottingham.ac.uk/
Received on Tuesday, 20 May 1997 13:17:52 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:26 UTC