Brian T Rice
Sun May 4 17:46:01 2003
Sorry for the time to reply. I spent some time thinking the situation
over before I developed a refined statement of the definition.
On Wed, 30 Apr 2003, Mario Blazevic wrote:
> Take your term "configuration", for example: "some collection of objects
> and relationships which hold over them" is a good start, but it doesn't
> say if the relationships are a property of the configuration or of the
> objects. To put it in a half-formal way, can you express a configuration
> as a pair C = (O: set of objects, R: set of relationships)? If you take
> the first component O of the pair, the set of objects, can you derive R
> from O without the configuration C?
I deliberately avoided stating "property"ship... TUNES is deliberately
abstract about core and derived ideas, and saying that "a is a property of
b" in a specification for it is overly-restrictive.
The closest analogue of "property"ship in the TUNES lexicon is
attribution, which is deliberately divorced from an ontology specifying
centrality of certain objects over others.
As far as the (O,R) formalization, I think this is decent, but needs
restating. It looks too much like "object-relational" however, which is
/not/ what is intended in the slightest. Really, you could formalize it
_Configuration_: a set of objects and meta-objects which is closed over
the operation of accessing the subject matter of any meta-objects in the
The objects closed over could be further meta-objects, which is why I
reduce it to one set. This begs the question, "how to distinguish objects
from meta-objects". My answer in Maude is that this is something
distinguished by the developer of a context. So it's a meta-programmable
thing, but not something that a machine can generate without some
inference system; there are no independent "first principles".
>From this definition, you cannot therefore derive the "relationships"
(the meta-objects) from the objects themselves. This is deliberately
dependent upon the situation (dare I use the term "context"? I am not
sure). So in the terms you state, the relationships (the meta-objects)
belong to the configuration on the configuration's own definition, and not
by the definition of the objects necessarily, which always involves the
creation context and always includes an unbounded number of irrelevant
By the TUNES HLL Principles (http://tunes.org/HLL/principles.html), under
Precision, we should have some scalable way of specifying what's required
out of a system to the extent of our interest or requirements. My
definition satisfies this by the "downward-closure" requirement from
meta-objects included to the objects which are their subject matter.
> Can you have two configurations C1 = (O, R1) and C2 = (O, R2) sharing
> the same objects but arranging them in different relationships?
However, your last question is pertinent in some sense. There can indeed
be two configurations with different meta-objects, and those differing
sets of meta-objects could even "express the same information" in
Brian T. Rice
LOGOS Research and Development