Re: First-Order Programming Theories
sowa <sowa@turing.pacss.binghamton.edu>
Date: Tue, 6 Jul 93 11:36:54 EDT
From: sowa <sowa@turing.pacss.binghamton.edu>
Message-id: <9307061536.AA05741@turing.pacss.binghamton.edu>
To: WALTER@cs.umn.edu, cg@cs.umn.edu
Subject: Re: First-Order Programming Theories
Cc: interlingua@ISI.EDU, sowa@turing.pacss.binghamton.edu
Walter,
Yes, the relativity of time is indeed a problem for distributed systems.
The authors of that book I cited did not address that issue.
However, the inability to have global time stamps does not preclude
the use of time as a sort. You can have local time stamps for each
context, which are not guaranteed to be comparable across contexts,
except when the time intervals exceed a certain epsilon, which would be
a function of the distance between them.
For example, you could have a time t1 in context 1 and a time t2 in
context 2. Both t1 and t2 would be of the same type, but if you tried
to compare them, you would get an answer unknown if they were too close
together.
In any case, there are certainly many questions about time that have
to be handled. But what I found attractive about that book is that
the authors presented a way of representing a very general programming
theory in FOL. The fact that they didn't solve all the problems of
distributed systems doesn't mean that such problems are insoluble
in some extension of their framework.
John