Path: utzoo!utgpu!water!watmath!clyde!rutgers!umd5!vrdxhq!dgis!csed-1!roskos
From: roskos@csed-1.UUCP (Eric Roskos)
Newsgroups: comp.os.minix
Subject: Re: porting minix to the mac
Summary: lack of low-level doc'n on Mac may be a problem
Message-ID: <254@csed-47.csed-1.UUCP>
Date: 21 Jan 88 02:50:52 GMT
References: <1436@lll-lcc.aRpA> <1825@botter.cs.vu.nl>
Organization: IDA, Alexandria, VA
Lines: 44

Another problem with porting to the Mac will probably be that the really
low-level interfaces are not well documented at all.  There is some boundary
between what's in ROM and what is loaded by the resource manager from the
system file, but (because Apple didn't want people to bypass the well-defined
interfaces) this doesn't seem to be documented anywhere in the amount of
detail needed to port another OS to the Mac.  For example, how system
initialization occurs (*in detail*) when you first start the system does not
seem to be documented anywhere.  This is likely to be a problem in at
least two areas (besides initialization):

1) There's no character generator, so ideally you would like to use
QuickDraw, probably, to write to the display.  But how to get QuickDraw
to use a particular font you have (given that the normal QuickDraw calls
use the Font Manager and Resource Manager) without initializing some other
parts of the Mac Toolbox may take some work to figure out.  (Also on some
of the earlier Macs there were not fonts in ROM.)

2) Disk I/O to the floppy disk drive is not done via a disk controller
the way it is on the PC; it is done via an "Integrated Woz Machine" which
seems to be just some simple control logic you can use to turn on the drive,
toggle the write line (or sense the read line) from the disk head, step the
head, etc.  They have a tight loop in the disk driver that sits and looks
at bits coming in, or generates bits going out, rather than issuing a
request to the disk drive like "read a block" or "seek to track n".  Again,
this means that the best way to do it would be to use the ROM, but again
it is not that well documented exactly how to do it.  (Also, while disk
I/O is going on you can't do anything else, which means you lose some
of your concurrency.)

This is not to say "don't try it," but rather just that the documentation
supplied by Apple is not as helpful as that provided with the PC (in the
Technical Reference) since they intentionally have tried to layer the
machine's architecture such that what you see as an applications programmer
is a "smart machine" with the underlying mechanisms hidden.  Actually this
is an admirable design goal (despite the many drawbacks of the conventional
parts of the OS, which are very primitive), but to keep people from
"cheating" they way they do on the PC (calling the BIOS directly, etc.)
I think an intentional attempt was made not to document the low-level
architecture.  There seem to be scattered "underground" documents where
people have gone in and figured out things by trial-and-error or
disassembly of parts of the ROM, though, (e.g., in MacTutor, maybe), so
with some research you probably can find those.  But it is likely to take
some time just to figure out how to work the machine before you can start
porting Minix to it.