RE: LANG: first sketch

I am less enthusiatic about the document than Michael Smith is.  I view it
as a proposal on how to think about generating a strawman, not the strawman
I was expecting.


The proposal divides the language into a ``frame-style part'' and a
``DAML+OIL/DL-style axiomatic part''.   This division is problematic for
two reasons:
1/ The proposal needs to give more information on what a ``frame-style
part'' of a language is.  The components of this part given in the proposal
include lots of notions that are generally not part of frame systems.
2/ DAML+OIL and other DL languages are very different with respect to the
concepts discussed in the proposal, and even DL languages differ
considerably on these concepts.  Which of the characteristics of these
languages are in this part of the language?

The proposal talks about determining which
``features/expressiveness/constructs'' the language should have.  This has
already been done.

The proposal says that syntax can be designed last.  Maybe it can, but
with no syntax, the other groups are going to have problems starting their
work.  It might be possible to leave details of the syntax until the end,
but some (preliminary) determinations on syntax will have to be done very
soon now.

The proposal suggests that the design should start with idioms.  This is
very likely to end up with a language with very many idiomatic expressions.
These idiomatic expressions will have associated names or phrases that each
will have connotations that will be understood slightly differently by
users who will often be surprised that the formal definitions of these
idiomatic expressions in the do not match up with their understandings.


I believe that a much better approach would be to take a large subset of
the requirements in the requirements document and build a strawman proposal
around them, including at least an abstract syntax and a sketch of a
semantics.  



Peter F. Patel-Schneider
Bell Labs Research

Received on Thursday, 7 March 2002 10:43:59 UTC