Path: utzoo!attcan!uunet!ncc!alberta!att!occrsh!uokmax!rmtodd
From: rmtodd@uokmax.UUCP (Richard Michael Todd)
Newsgroups: comp.os.minix
Subject: Re: Status of V1.3
Summary: What's so great about Version 7?
Message-ID: <1302@uokmax.UUCP>
Date: 13 May 88 01:31:46 GMT
References: <709@ast.cs.vu.nl> <4530@hoptoad.uucp>
Reply-To: rmtodd@uokmax.UUCP (Richard Michael Todd)
Organization: University of Oklahoma, Norman, OK
Lines: 66

In article <4530@hoptoad.uucp> gnu@hoptoad.uucp (John Gilmore) writes:
>The compress documentation should say that it's a 12-bit compress (if it is),
>not the standard 16-bit compress.

If the compress he's got is the one I posted back when I posted the PD Tar
changes, it's a 13-bit compress, not 12 bit.  You can cram 13-bit compress
into 64K split I&D and have it work.  

>A real version of tar(1) should be included, e.g. one that uses the same
>arguments as v7 tar.  Mine is in comp.sources.unix.

And it's already been ported to Minix, too.  The original Minix tar didn't
even output tarfiles to stdout--meaning you couldn't use "tar c ... | bundle"
to backup your hard disk directories easily.  With PD Tar you can; that's
one of the main reasons why I ported it.  

>Has anyone retrofitted the Berkeley TCP/IP networking code recently posted? :-)
>Or even Phil Karn's stuff, which already runs on MSDOS IBM PC's?
Won't be terribly easy to do, if it's even possible.  (BTW, if I'm not
mistaken Karn's stuff contains much of the Berkeley TCP/IP code, just the
drivers are different.)  The main problem is limited address space--the
original BSD code was written on Vaxen where you can allocate plenty of
room to mbufs.  Keith Bostic et al., when they added the Berkeley networking
code to BSD2.10 (the PDP version of BSD), had to do all sorts of nasty
mapping segments in and out to get it to run in the PDP's address space
(roughly like the MINIX system of 64K I, 64K D space for each process).
You'd almost have to add another system process (like FS and MM) just to
handle the network.  

>Bring the standard library up to snuff so that it can compile the
>average 1988 program posted to the net, e.g. all the string functions,
>(Henry Spencer in comp.sources.unix again), both BSD and SV
>bcopy/memcpy, the Posix directory access routines, etc.  Reading the

Just a reminder, I've already done a port of the Posix directory library
to Minix.  Didn't really need much changing to get it to work.

>Porting my tar to Minix involved writing execlp(), bcmp(), bzero(),
>3 arg open.
Well, 3 arg open didn't *have* to be written, since V7 systems got along
without it, but it did allow using the full functionality of the PD Tar
(specifically -k).  I also had to revise printf() to allow both the %ld
formats that the rest of the planet has been using for the past ten
years :-), as well as variable-length formats.
  Ideally I should be able to take a program off of the net and, assuming
it isn't using some esoteric feature like SysV IPC or BSD sockets, get it
running under Minix with as little trouble as it takes to get it running
on one of our Suns or the Multimax.  Reinventing library routines is not
my favorite activity.  

>Small may be beautiful, but small and broken is ugly.
He who thinks small is beautiful should see my dorm room for a
different opinion :-).  
  Seriously, although a certain amount of smallness is required by the
limitations of the XT hardware, we could theoretically do a lot better.
I hold up Berkeley Unix version 2.10 as an example.  The authors of 2.10
have managed to get a reasonable subset of full BSD4.3 to run on PDP-11
machines.  PDP-11's aren't that much bigger than XT's, and they share 
many of the same address space limitations.  We should at least be able
to get some of the Berkeley features working on our machines.  So why
should we limit ourselves by compatibility with Version 7, a Unix variant
that's about 10 years behind the times?
-- 
Richard Todd		Dubious Domain: rmtodd@uokmax.ecn.uoknor.edu
USSnail:820 Annie Court,Norman OK 73069 	Fido:1:147/1
UUCP: {many AT&T sites}!occrsh!uokmax!rmtodd