COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc]" source
C00006 ENDMK
C⊗;
.require "memo.pub[let,jmc]" source;

Referee's report on %2Mechanizing Temporal Knowledge%1 by Kenneth Kahn
and G. Anthony Gorry

	The value of this paper as a contribution to artificial
intelligence depends on an issue not discussed by its writers -
namely the extent to which a problem involving time and some
other facet of reality, say space or the ownership of property
can be divided into parts one of which is the time specialist
and the others of which - including any overall co-ordinator -
don't know about time.  My own opinion is that this can't be
done - and therefore the time specialist isn't of much use -
but I don't want to impose this opinion on the authors or the
readers of the ⊗Journal.  It seems to me that some facts about
time can be stored separately from others and co-ordinated
with facts about other matters by a general problem solver
of some kind, but that the utility of a specialist is doubtful.
Even then, some of the most important facts involve both time
and other topics, e.g. facts about motion.

	I think the authors should be required to disuss this
issue at least briefly.  Moreover, they and other authors of
papers submitted to the ⊗Journal should be required to explicitly
discuss the limitations of the methods they use in the paper and
the limitations of the methods they use.  Too often papers are
written so as to sweep under the rug the limitations of the methods.
For example, this paper ignores events extended in time and the
notion that one event occurs ⊗during another.  We have the
sentences

	1. %2John has visited Paris twice%1.

	2. %2John has been visiting Paris for three weeks%1.

	3. %2John was going to school when he fell off his bike%1.

	4. %2John came every other week%1.

Here are some other comments:

1. Fig. 2 shows no output from the user or input to memory.

2. The statement on p. 11 that "Typically, the time of a special
reference event is known quite precisely" seems dubious optimism
about the applicability of the program.

3. The time travel example seems a strange way to go given the
inability of the system to deal with the more important notion
of duration.

	On the whole, I think the paper should be published if the
authors discuss the issues mentioned above.