[tmql-wg] TMQL Use Cases: Remaining Issues

Robert Barta rho@bigpond.net.au
Tue, 6 Apr 2004 18:26:15 +1000

Hi all,

I had a look at my left-over notes for the outstanding issues for the
TMQL use cases document. Some of these I have collected without
understanding their merit, so I am simply listing them here.

Some issues would involve that new use cases have to be added or -
more drastically - some of the data will have to be changed. As I do
not want to make such changes unilaterally I will post suggestions
here and only if there are no major objections, will add them to the
document.

I will certainly not add anything before Amsterdam.

\rho

==================

From Lars:

> The subclass association does not use the right PSIs, so this
> means that query engines can't actually know that this is the
> magic superclass-subclass association type.

Correct.

Currently, we have in the text:

"It is assumed that subclasses represents
http://www.topicmaps.org/xtm/1.0/core.xtm#superclass-subclass,
superclass represents
http://www.topicmaps.org/xtm/1.0/core.xtm#superclass and subclass
represents http://www.topicmaps.org/xtm/1.0/core.xtm#subclass.

This association type should have the implicit property that if a is
an instance of A and A is a subclass of B then a is also an instance
of B. The binary relation subclasses is to be interpreted
non-reflexive and transitive."

? Should we change the data to reflect this statement ?

From Lars:

> uc-literature.xtm is not valid XTM, since <subjectIdentity>
> appears at the end; it has to go between <instanceOf> and
> <baseName>. (Found this in the new validating Omnigator.)

This is now fixed in the CVS version.

From Lars:

> The sort names are not actually sort names:
>  1) they are not variant names, and
>  2) they do not use the correct PSIs (or any PSIs at all).
>
> It's still possible to sort on them, but they should be actual
> sort names.

I wonder, why they should, so what difference it makes.

In the prolog for the use case data we have currently:

"the basename value in the scope <code>sort</code> we call the
<emph>sort name</emph>"

? Any good reason to change this?

From Lars:
> The sample data does not have any subclass of tutorial with
> instances, so the "direct or indirect" part does not have any
> effect.  I guess we should either take it out of the use case or
> add data to the topic map.

True. We have 5.2.2.8, though, which asks for this.

From Lars:
>  "publications" should here probably be "documents", since that's
>  the topic type.

Not quite. Currently, the accompanying text says:

"Conversely, a publication is a document which has been authored
by someone."

So there must be an authorship relation to make it a publication.
(Should actually be expressed in TMCL :-)

From Lars:
> The results shown include non-papers TAO and colourful.

True. Fixed in CVS.

- New Use Case

Should we have a use case covering more complex associations with
several roles? All associations currently are binary.

If so, then the data has to be changed. Idea:

(paper-is-presented-at)
conference: xml200x
paper: some-paper
presenter: someone

Use case:

"Return all titles of documents which have been presented at the xml200x
conference."

- New Use Case

Do we want one for "types for basenames"?

Idea:

?????

- Use Case for reification of topics

Actually 5.2.2.16 should be read like this:

"Which topics reify something on that site"

I see that all solutions have interpreted it like this.

- Use Case for reification of associations

Change in the data:

(paper-is-presented-at) is-reified-by x123
conference: ....
paper: ....
presenter: ...

x123
bn: ...
in (entertainment-level): 5

Use case:

"return all papers which were presented with an entertainment
level 5 (and above)"

- Use Case for reification of characteristics (yuck)

Change of data:

x123
in (entertainment-level) is-reified-by x234 : 5

(is-assessed-by)
assessor: xml200x-audience
assessment: x234

Use case:

"return all presentations (paper and presenter) which
have been assessed by the XML200x conference audience"

- Use Case for source locators

????

- Use Case for Full Text Retrieval

Do we need more sophistication than what we have in 5.2.2.13?

Soundex, ranking :-)

===

- There references were missing. I have regenerated them and committed
a new version into the CVS. Maybe the Makefile does not work when being
executed by Lars?

- In the visible version at

http://www.isotopicmaps.org/tmql/use-cases.html

most links do not work (documents not copied). Also the small image
at the beginning does not show up. It does in the CVS version.

- Do we need a "Conformance Test Suite"? It would formalize the use cases
and combine them with the expected outcome (CXTM?)

I think this should be done as soon as we have a bit more machine support.
Maintaining the use cases manually is VERY error prone.

- Shall we add here cases to outline that processors are allowed to
do 'lazy evaluation'? This is also something which might have to
be reflected in the TMQL requirements document.