[ba-ohs-talk] Fw: A modest proposal
> Now, here's the crucial revelation. What happens when you put the same
> topic into two different associations? (01)
Sorry, that should have read, "same topic into two separate associations
with different scopes". (02)
> The resulting XTM 2 would be able to do anything that DAML-RDF could
do, (03)
Actually not quite.
I would change the <instanceOf> relation in an association to
<signifier>.
The relation is not really class-instance, it is signifier-signified.
The association is the signified of the signifying topic. (04)
Then if you've removed the internal scoping of topics as I've suggested,
the possibility of creating
association templates becomes a reasonably simple matter of using the
instanceOf
relation between the
two signifying topics. (05)
Then it would be as powerful as DAML-RDF. (06)
Enjoy. (07)
--
Peter (08)
----- Original Message -----
From: "Peter Jones" <ppj@concept67.fsnet.co.uk>
To: "Jack Park" <jackpark@thinkalong.com>; "Murray Altheim"
<m.altheim@open.ac.uk>
Sent: Thursday, March 21, 2002 7:52 PM
Subject: A modest proposal (09)
> Guys,
>
> I need to float a _serious_ idea in front of you two.
> I'm hoping that you'll go and float it in front of some other folks
> whose
> acquaintance we have in common.
>
> I've just had a revelation. First, an assertion:
> GSIX and XTM are not isomorphic.
> I'm confident of the truth of that (see below).
> This is because scopes in TMs and contexts in GSIX are different.
> However,...
> I've just figured out _precisely_ what that difference is.
> (Background: I started GSIX because I was unhappy about the internal
> topic structure of TMs and the limiting effects of that.
> I only saw it intuitively at first, and I was hoping that GSIX would
> trigger some more powerful whizz to fill in the gaps.
> But perseverance has provided me with valuable learning and maybe now
I
> can put all of that to the service of higher aims than a
> single spec.
> It's my belief that that internal topic structure is the single, and
if
> what I'm about to explain is correct, only 'flaw'* of TMs in respect
of
> the lofty interchange goals XTM set out to achieve.
>
> Let's assume, for the purposes of my exposition, that you have some
> topics whose only properties are an DUID and a baseName. No internal
> topic scoping.
> Then let's assume that a topic doesn't go 'live' in a topic map until
it
> becomes an association member.
> Then assume that when a topic becomes a member of an association, the
> topic has no other scope than that of the association. It inherits its
> scope from the association.
> Now, here's the crucial revelation. What happens when you put the same
> topic into two different associations?
> In XTM 1.0, the spec says it's the same topic which you then give two
> different scopes to.
> And then the whole scoping properties regress rears it's ugly head an
> you have to limit scopes to topic IDs instead of having
> rich, deep contexts.
> The answer, as I see it, and as I have done my best to assert in GSIX,
> is that the same baseName under two different scopes should become
> two separate topic objects with separate IDs. Then scopes can become
> contexts with depth without fear of regress.
> How do you assert that they are the same topic, semantically?
> One way is to have a scope independent indexing mechanism (also
> something I've shoved into GSIX, possibly rather badly as yet).
> The other is to bind them to a higher common identifier, and perhaps
the
> mechanisms in TMs for that are already there.
>
> Now, where am I going with this? You've probably already guessed.
> GSIXTM 2, or something similar. Probably more XTM 2 than GSIX since
with
> the modifications above the XTM syntax starts to look friendlier than
> GSIX.
>
> * One issue to discuss. Is the 'shallow' or 'flat' scoping of ISO13250
> and XTM1.0 a flaw? It could be considered beneficial if processing
> cycles were at a premium. But it limits the applicability of TMs,
imho,
> and I would forecast that Moore's Law will take care of that issue.
>
> The resulting XTM 2 would be able to do anything that DAML-RDF could
do,
> and more since it wouldn't be tied to the ontology they've adopted,
and
> would improve upon GXL because GXL holds useful content in attributes,
> which kills i18n amongst other things.
>
> What say ye?
>
> Cheers,
> --
> Peter Jones
> ppj@concept67.fsnet.co.uk
>
> (010)