RE: type safety

• To: "'types@cis.upenn.edu'" <types@cis.upenn.edu>
• Subject: RE: type safety
• From: Shriram Krishnamurthi <shriram@cs.rice.edu>
• Date: Sun, 2 Jan 2000 19:26:43 -0600 (CST)
• In-Reply-To: <200001021912.OAA14226@saul.cis.upenn.edu>
• References: <200001021912.OAA14226@saul.cis.upenn.edu>
• Reply-To: shriram@cs.rice.edu

Greg Morrisett wrote:

>		       I've found that the soft-typing approach
> for Scheme doesn't work well unless you program in a very
> stylized fashion -- in particular, you must use Andrew Wright's
> wonderful pattern matching system to get decent results.  

As someone who has used soft typers extensively, I must disagree with
this contention.  What you say is true of Andrew Wright's Soft Scheme,
which was implemented to exploit patterns.  Our experiences using Soft
Scheme on realistic software, which had the flavor you describe,
affected the design of Cormac Flanagan's MrSpidey.  MrSpidey doesn't
depend on patterns; indeed, it macro-expands patterns away, so using
patterns would give you no appreciable advantage.

Based on your comment, I'm guessing you've used Soft Scheme (ca. 1993)
but not MrSpidey.  I don't speak for Matthias and Corky, but I expect
that when they talk about what soft typing is capable of, they have
both systems in mind.  Soft Scheme was an interesting generation in
this lineage, but is by no means the latest one.

'shriram

• Prev: RE: type safety
• Next: Re: abstraction power
• Index(es):
  • Main
  • Thread