[sc34wg3] revised draft Reference Model document N0298

Jan Algermissen sc34wg3@isotopicmaps.org
Tue, 09 Apr 2002 15:45:26 +0200


Steve, Michel--

thank you for your very kind words here and in the acknowledgement and
there is really no need to apologize!

I have been waiting for the revised version to make my comments and to
ask my questions, here they are:

I am impressed!

You really destilled the fundamental ideas into a very simple model
with great potential and into a very readable document (which was
the hardest part I suppose ;-) .


I think by making subject identity an assertion itself you found about
the only solution for the problem that multiple text elements such as
<baseNameString> can in fact *indicate* one and the same subject. When
I read your explanation regarding literals I (finally) understood that
things like the following are possible:

Suppose you have the markup

<integer id="i1" base="16">ABCD</integer>
....
<integer id="i2" base="10">43981</integer>

There are two addressable subjects: S1: the string found
inside the <integer> element with the ID i1 and S2: the string
found inside the <integer> element with the ID i2.
Suppose my application 'knows' the semantics of the <integer> element,
I can then have another subject S3: 'the integer 43981'
By making S1 and S2 subject indicators for S3, the dRM allows to
keep all the different notations in the topic map, making them
available for display for example.


Now questions:

Will the reference model explicitly name/define the roles that 'belong' to the
two assertion patterns (role-pattern, role-role, role-rolePlayerConstraint,
role-topic and role-subjectIndicator) ?


By making subject identity an assertion, it is now possible to 'reason' about
the connection between a node and it's subject indicator. Especially it will be possible,
at the SAM level, to scope that assertion, thus limiting its 'extend of validity'.
Do you see any negative implications regarding merging ? ( I think that I am mixing up
dRM and SAM level but I am asking anyway )


Suppose there are several players of the same role in an assertion, will this be represented as
several ACRx subgraphs ? Or will the x-node represent the *set* of topics playing that role ?


Jan


-- 
Jan Algermissen
Consultant & Programmer
Mobil: 0177 / 283 1440
www: http://www.topicmapping.com