Path: utzoo!attcan!uunet!snorkelwacker!mit-eddie!uw-beaver!uw-june!jmaloney
From: jmaloney@cs.washington.edu (John Maloney)
Newsgroups: comp.object
Subject: Re: code blocks (aka closures)
Message-ID: <12448@june.cs.washington.edu>
Date: 2 Jul 90 23:25:29 GMT
Organization: U of Washington, Computer Science, Seattle
Lines: 22

Brad Cox and Andrew Koenig's messages nicely summarize both
the motivation for and the implementation problems with adding
closures ("blocks") to a language like C++ or Objective-C. He
points out that the implementation problems are deep because:

    full closures require dynamically allocated variable contexts,
    a garbage collector is required to get rid of these contexts when
        they are no longer referenced, and
    adding garbage collection to C is hard because it is difficult
        for the runtime system to know for certain which variables
        contain pointers and which contain integers.

Syntax and parsing issues are relatively unimportant compared
to these implementation problems. I still maintain that closures
are a powerful and useful language feature and think they
should stay on the "wish list". Of course, I also believe that
garbage collection is well-worth the 10% overhead that a good
incremental garbage collector consumes. Programmers' time
and program correctness are far more  important than a
program than runs a mere 10% faster.

	-- John