Re: SUO: Re: CG: Negotiation Instead of Legislation
Well, let me complement your reasons with an overview of what
the Mozart system does in those areas.
> There are three reasons for picking BinProlog:
>
> 1. It has a highly optimized compiler that generates
> either C or Java byte codes (the Java version is
> comparable to the best interpreted versions of Prolog
> and the C version is much faster).
The Mozart system compiles to its own efficient byte code. (It
cannot compile to efficient Java byte code because of the
higher-order, symbolic, concurrency, and distribution features,
which are poorly supported by Java byte code.) The advantage is
that Mozart has a distribution subsystem that permits transparent
interoperability between Windows and various flavors of Unix.
There is of course a performance hit for programs written in Oz
when compared to programs written in C. This can be compensated by
using the Mozart C++ interface, which is quite sophisticated (e.g.,
the different built-in constraint systems use it), to implement
low-level algorithms that *must* be fast. As we all know, global
performance depends on only a small fraction of any program.
> 2. Paul Tarau, the BinProlog developer, has been working
> with Arun Majumdar to implement features to support
> CGs directly and to tune the underlying implementation
> to optimize the CG algorithms.
As far as I know, there has been no CG work using Mozart (but I could
be mistaken). There is a large effort in computational linguistics
in Mozart, though, by the people at the University of the Saarland,
in Saarbrucken, Germany. There are various projects in discrete
optimization, agent platforms, parallel simulation, collaborative
applications, and human-computer interfacing. I can give references
to these.
> 3. The higher-order, modal, metalevel, and contextual
> features are supported by CGs. Any reasoning that
> would use those features could be done in CGs instead
> of the implementation language.
Implementation of higher-order and contextual features may be easier
in a language that supports well higher-order programming and full
compositionality.
> I put Paul Tarau and Arun Majumdar on the cc list for this
> note. Perhaps they would be better qualified to comment
> on these issues.
>
> John Sowa
Good idea -- I'm sure that Paul has something to say!
Peter
--
Peter Van Roy
Département d'Ingénierie Informatique
(Department of Computing Science and Engineering)
Université catholique de Louvain
B-1348 Louvain-la-Neuve, Belgium
Email: pvr@info.ucl.ac.be
Tel: (+32) (10) 47.83.74
Web: http://www.info.ucl.ac.be/people/cvvanroy.html
Mozart: http://www.mozart-oz.org