Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site steinmetz.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxt!houxm!whuxl!whuxlm!akgua!mcnc!ncsu!uvacs!edison!steinmetz!davidsen
From: davidsen@steinmetz.UUCP (Davidsen)
Newsgroups: net.micro.pc
Subject: Re: PC/IX
Message-ID: <339@steinmetz.UUCP>
Date: Wed, 11-Dec-85 16:07:52 EST
Article-I.D.: steinmet.339
Posted: Wed Dec 11 16:07:52 1985
Date-Received: Sun, 15-Dec-85 00:23:15 EST
References: <2300@sdcc6.UUCP> <964@homxb.UUCP>
Reply-To: davidsen@kbsvax.UUCP (Davidsen)
Organization: GE CRD, Schenectady, NY
Lines: 146
Summary: PC/IX experience


I have been running PC/IX since early 1984. First on an XT, then a beta test
version for the AT, then release 1.1 from IBM. I have also had a chance to
test XENIX 1.0 from IBM. Since I have had a chance to run the two systems on
two AT's, I have had a perfect chance to compare the systems. I have a copy of
SCO XENIX 5.0 on order. For those who want the bottom line, with a copy of
each system in hand, I'm running PC/IX on my personal AT system.

Some comparisons:

XENIX is *much* faster at nroff/troff. In fact, the 6Mhz AT runs about as fast
as a lightly loaded VAX 11/780 in that respect. XENIX has made some
"enhancements" to the macro packages (our local expert says particularly mm)
which can lead to developing input text which won't pass through other
systems. This is reason one for my choice of PC/IX.

XENIX uses all the memory in the machine, but allows only 64k for any given
chunk (array or whatever).  Another interesting gotcha is that static arrays
may be 64k, but calloc() will only go to 32k (I suspect signed arithmetic in
there somewhere).

The XENIX compiler runs slower than the pcc compiler in PC/IX, but not much.
The compiled programs run considerably faster, and you can compile with a DOS
option and produce a DOS version of the program. The gotcha here is that the C
compiler is *not* pcc, and on my personal benchmark suite (developed over 21
years as a programmer/software engineer/consultant, etc) fully 1/3 of the
programs either did not compile or did not run to completion. This suite has
been run on everything from PC's to a Cray2, V7, Sys3, SysV, BSD4.2, etc,
without changes, but failed miserably under XENIX. This is reason two for my
personal choice not to run XENIX.

Configuration of PC/IX is very versatile for things like ports which autobaud
at strange baudrates, such as 9600, 1200, 300. XENIX allows groups for the
lower speeds, but doesn't seem to have the ability to pick a group of speeds
and parity selections. The selection in PC/IX is done using a text file with
entries such as "speed=9600,300,1200", which is easier to remember than the
coded entries in XENIX. Reason three for PC/IX, but I realize that most people
don't have this problem, and therefore the solution is not of value.

PC/IX gives reasonably explicit directions for writing new dialers. They
provide dialers for a number of common modems, but allow easy user
installation of new dialers. Although most people will not have to support
three or four new modems (bet you wish you did), the new 2400 and 9600 baud
modems require handling of additional status messages, etc, back from the
modem. Another reason for me to go with PC/IX.

PC/IX also allows easy installation of printer filters, including multiple
modes in the same physical printer being treated as separate printers. I
haven't had to do this yet in XENIX, and have no idea what is entailed. The
queueing system also allows producing queues for troff, format and uucp for
remote printing, etc. It is quite versatile and easy to use, and most of the
setup is done by setting cleartext parameters in system files.

Although neither o/s provides support for additional serial ports from IBM,
XENIX drivers are available for the XT if you look hard, and readily for the
AT. I hope to try the MA systems 2.5 MB +2 serial in 1986. This is just not
available in any way for PC/IX.

XENIX requires a minimum of 15MB on a 20MB disk in an AT, while PC/IX may be
run in as little as 7MB for a complete system, or 3MB if you just want to
run applications and mail. XENIX (as from IBM) won't work properly on a 32MB
hard disk. It installs, won't use more than 20MB of the disk, and fails to
boot with a CRC in a non-existant sector. This was on three drives tried in
two machines. If booted from floppy, the 20MB system runs fine, but takes
three partitions of the disk. PC/IX has worked on every disk size tried to
date, including 20 and 32MB, and a few oddballs. I have seen patches from CORE
(as I recall) which allow use of other disk sizes. Installation is *not*
trivial, and the instructions were for one odd disk size with no general info
on how to do disk X.

At power-up, XENIX requires two keystrokes to put it into multi-user mode.
This means that if a power failure occurs, the system must be visited by a
live human being. PC/IX checks the file system and goes into multiuser mode by
default (although this may be modified). This gives recovery after power
failure and also allows system startup by people you would *never* let make a
decision, such as young children and computer haters; "see the red switch on
the rihgt side? push it UP! Thank you". One PC/IX system was inadvertently
cycled five times by electricians, and rebooted correctly each time. The XENIX
system on the same feed required manual intervention to recover the file
system (but did not lose any useful data).

XENIX has all of the standard archival programs: tar, cpio, and dump/restore
(renamed to backup/restore). PC/IX is notably missing tar. XENIX also has vi
and more (the program more), although PC/IX has INed and l, and both more and
less have been posted to net.sources, so the features may be user installed.
The vi editor is more standard, but INed is perhaps a bit easier to use (this
is more a matter of religion than a technical issue). INed can not be used
from a remote terminal (a marketing decision, the version on a VAX can) while
vi will work remotely.

For all the ease of configuration of PC/IX, the XENIX method, while harder to
use and less flexible, looks closer to my generic Sys3 documentation. The
XENIX manuals are easier to read, but somehow always leave me looking for just
a bit more detail. The PC/IX manuals are standard UNIX manuals and are best
described as "concise". The XENIX manuals win hands down for the starting
user.

The installation of PC/IX is a snap. It takes only one partition, of any size,
does not need to be any particular partition, and can be run without the
manual by those able to read the menus and questions. I have installed
XENIX at least eight times in the last six months, and finally wrote a cheat
sheet to avoid having to read the manual every time. This is one place where I
favor menus rather than commands, since a complete installation is (hopefully)
not something you want to do often enough to learn by heart.

Finally, if you have to live with the DOS world, the file transfer to/from DOS
disks and partitions works well in PC/IX, with the exception of singlesided
DOS disks, which don't seem to write correctly. XENIX has given me "learning
experiences" with a number of floppy formats, and the one time I tried to
write the DOS partition of the hard disk, I had to reformat the partition. The
XENIX commands for DOS interface seem somewhat better chosen than those in
PC/IX, but neither is a joy to use.

XENIX offers the visual shell (vsh) and csh in addition to sh. PC/IX has only
sh, but offers TEN+(tm?) as an option (for more money) which seems to be
slightly better. There is some form of vsh alailable in public domain, for
those who like that sort of thing (another religious matter). Finally, the
price of the complete systems is about the same, but the PC/IX system is not
unbundled, while XENIX is available as three parts: UNIX, C, and word
processing.

Power: for the power hungry single user, the AT with an 80287 (an perhaps a
fast clock) is a good alternative to sharing a mini. Every benchmark I have
run, both my own and such things as Dhrystone, have indicated that an AT at
8MHz is faster than a VAX 11/750 in every CPU category, and although slower in
disk access, is still faster in total time for things like compilation, nroff,
awk, grep, etc. No version of UNIX for the AT currently supports the "huge
model" with objects > 64k, but I am told that this will happen in XENIX
eventually. XENIX runs in protected mode, while PC/IX runs in real mode. For
the person using UNIX tools for software development or applications there is
little difference in the performance of the two systems. The larger memory of
XENIX may be offset by the non pcc compiler. FORTRAN is available for both
systems, but since I have no XENIX version I will not comment.

---------------------------------------------------------------- 

Disclamer: these are my personal findings and opinions after several years of
running UNIX on PC's. My work with VENIX and Coherent has been too casual to
report, and there may be some non-obvious features of either or both systems
which I have not used or reported. These finding are mine alone, and do not
reflect the findings or opinion(s) of the General Electric Company.
-- 
	billD	(..seismo!rochester!steinmetz!crdos1!davidsen)
		(davidsen@GE-CRD.ARPA)

"It seemed like a good idea at the time..."