Re: data types and initiality

To: types@theory.LCS.MIT.EDU
Subject: Re: data types and initiality
From: thatte%cs@ucsd.edu (Satish Thatte)
Date: Thu, 10 Nov 88 11:36:59 EST
Sender: meyer@theory.LCS.MIT.EDU

Date: Thu, 10 Nov 88 08:02:09 PST
To: jcm%ra.stanford.edu@stork.lcs.mit.edu
Cc: types@theory.LCS.MIT.EDU

I think the main argument in favor of the initiality point of view is
its universal applicability.  The "observational" (i.e., final) models
need to assume some form of the "sufficient completeness" condition
of Guttag-Horning wrt the basis for observation, in order to guarantee
their existence.  It is possible to give a universally applicable
final model, but then one needs to make restrictive assumptions about
possible extensions of the specification, etc.  In other words, such
a model is applicable to *languages* of specifications rather than to
individual specifications.  I am in the process of writing a paper
on this approach with Larry Moss at IBM Yorktown Heights.  I can
send you a copy if you are interested.

Satish Thatte