Re: MzScheme vs MIT Scheme

[Matthias Felleisen <matthias@cs.rice.edu>]

MIT Scheme is older than MzScheme. It adheres to the philosophy of MIT
Scheme researchers, which says that almost everything should be accesible
to programmers (e.g., environments, closures, etc). 

MzScheme+ and MIT Scheme-

 * units (aka modules for components)
 * first-class classes (aka "object system")
 * generative structrues
 * exceptions 

MIT Scheme+ and MzScheme- 

 * control over run-time from within programs 
 * OS/2 port 
 * older and therefore more traditional libraries around

Both support r^5rs macros to some extent. In MzScheme, you load the library
synrules.ss and get a reasonably decent implementation of these macros. The 
port is due to Dorai Sitaram (GTE Labs) and Shriram Krishnamurthi (Rice,
still). 

;; --- 

The size of DrScheme is a small problem though I am running it on a similar
machine (right now) and experience no such size problems. 

Why is DrScheme so large (11Mb at start-up, 20 within a few runs)? 

 The best answer is "because you get so much more from DrScheme than from
 an ordinary Scheme repl". It gives you access and control over a huge GUI
 library, which basically means you get an OS.

 The worst answer is "because we use Hans Boehm's `conservative' garbage
 collector." It holds on to far more memory than necessary. Also, we
 haven't figure out a way to return pages to the OS on a cross-platform
 basis. 

There are several answers in-between those two. I can explain what DrScheme
provides in detail, too. 

Still, my experience is that for small programs (< 1000 loc) DrScheme poses
no memory problem. I develop them routinely in there, using two or three
windows. For larger programs, Robby should comment; he can [could?] run
DrScheme on DrScheme. 

-- Matthias