Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ukma!rutgers!att!lzaz!bds
From: bds@lzaz.ATT.COM (B.SZABLAK)
Newsgroups: comp.os.minix
Subject: Re: Bug in fs/device.c, ioctl.c, tty.c
Summary: MINIX programming philosophy/style
Message-ID: <499@lzaz.ATT.COM>
Date: 28 Mar 89 14:05:29 GMT
References: <11714@louie.udel.EDU>
Organization: AT&T ISL Lincroft NJ USA
Lines: 14

In article <11714@louie.udel.EDU>, des@berlioz.nsc.com (Desmond Young) writes:
>   2. The person who added the TTY_SPEED assignment defined it as m2.i3
>      (3rd int of message type 2). THIS IS THE FIELD USED TO STORE THE PROC
>      NUMBER (OH SHIT)...

My biggest complaint about the way MINIX is programmed is this overloading
of message fields. It is confusing and error prone (at least when modifing
things. I would have prefered a more "natural" approach:

struct { int msgtype; union { struct type1 m1; ...; struct typeN mN; }; };

I can appreciate the problems with such an approach, but for an
"instructional" OS, I believe this is a better way of doing things
since it is clearer and less error prone (a good thing even for "real" OSs).