Subject: Re: Emacs and CL (was Re: More LispOS talk (was Re: Lisp subverts the world (was Re: ints vs fixnums (was Re: Java ... (was Re: ... (was Re: ...))))))
From: Erik Naggum <>
Date: 1999/03/03
Newsgroups: comp.lang.lisp
Message-ID: <>

* Tim Bradshaw <>
| What I meant was that it's possible to share some static parts of the
| heap, either by lazy copying (the good way) or by the emacs trick.
| per-process data is obviously not sharable.

  then this needs to be _added_ to GUILE.

| However, at least in theory that per-process data should not be
| significantly more than the per-`application' data in a shared
| address-space system.

  this theory does not apply to GUILE.  GUILE consumes a _lot_ of memory.

| As to how Guile is meant to be used, I don't know about that.  I was just
| trying to argue that the Unix model doesn't have to be death in theory
| even though it may be in practice.

  I think it's death in theory, but doesn't have to be in practice, the
  same way whole populations don't die out under dictatorships or under
  oppressive taxation -- i.e., the population survives because people do
  the _wrong_ thing according to its planners and designers.

  it has been said that Microsoft's "operating systems" are nothing but a
  boot loader with a GUI.  the process model in Unix is likewise geared
  towards a mode of operation that may cause it to be trivialized the same
  way, in not too distant a future, if it has not already happened.

  think about it, the programming environment model in Unix is an imitation
  of the Lisp machine, and the way it is implemented is through processes
  and inter-process communication instead of function calls.  functions in
  the Lisp heap are programs on disk in the Unix model.  the optimizations
  that made Unix able to survive this incredible inefficiency are bad for
  Lisp, which would have done it a lot better had it been in control, but
  GUILE doesn't do it any better.  GUILE combines the worst of both worlds,
  in an attempt to bring the best of one world to a world where it doesn't
  really belong.  on top of this, it's Scheme, and it's expensive to run,
  there's so much wrong with it that nobody can use it without fixing some
  part of it.  so, it's going to be a winner, but a real Lisp world would
  be so much better.  however, a real Lisp world cannot win as long as
  GUILE is the image people have of a Lisp environment, just like Lisp has
  suffered tremendously from the Scheme people's insistence that Scheme as
  taught to college students is what they should expect from Lisp.

  I'm not sure what needs to be done about this dismal situation, however.