[sc34wg3] Some comments on TMRM v6.0

[Bernard Vatant]

Steve, Ptrick and al.

Some first and very partial comments at first reading, I had not the bandwidth to plough
through the whole document, and the introduction of the T+ model makes it, for me at
least, much more difficult to parse than the previous versions. Gone only through Section
2 so far.

- I had hard time trying to make sense of the definitions on page 4. Seems to me that
there are some inconsistencies in the use of notations.

X is defined as a set of proxies, but it's unclear if it's the set of all (possible)
proxies in the TM universe (so to speak), or all proxies *in a given map*.

If k is a key and v its value, it's asserted that (k,v) is element of (X ? V). This seems
clearly inconsistent with the definition of X being a set of proxies. Moreover V itself is
not defined, although one can guess it's a set of values.

This does not help to understand those primitives. I really would like to catch the point
of defining a property as a couple (key, value), but so far I miss it, and the following
comment "Properties are labeled values" adds to my confusion. I have some notion of what a
value can be, of what a label can be, maybe less on what a key can be ... but what is the
relationship between this label and that key? Same thing, different? I'm lost, so I tried
on the examples

Proxy identifier 	property 	value 	property 	value
aaa 			name 		Durusau 	SSN 		1231

I guess "name" and "SSN" look like "labels" of the properties. So, what are the keys?

- I'm amazed that TMRM mingles in definition of built-in proxies for class, instance,
superclass and subclass. Those IMO should be defined by disclosures, since there is no
universal definition neither of those subjects nor of the rules governing them, all of
them depending of the flavor of KR/logic framework you choose. I always had actually the
same concern with similar built-in PSIs in XTM 1.0, and I fail to see why TMRM needs
definition of such subjects anyway.

That's all for the moment

Have fun in Montreal

Bernard

----------------------------------
Bernard Vatant
Mondeca Knowledge Engineering
bernard.vatant@mondeca.com
(+33) 0871 488 459

http://www.mondeca.com
http://universimmedia.blogspot.com
----------------------------------