Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site watdcsu.UUCP
Path: utzoo!watmath!watdcsu!herbie
From: herbie@watdcsu.UUCP (Herb Chong, Computing Services)
Newsgroups: net.followup
Subject: Re: VM/IX on IBM 4341 (some views from the other side)
Message-ID: <722@watdcsu.UUCP>
Date: Fri, 30-Nov-84 21:10:52 EST
Article-I.D.: watdcsu.722
Posted: Fri Nov 30 21:10:52 1984
Date-Received: Sat, 1-Dec-84 19:40:47 EST
References: <117@circadia.UUCP> <411@mhuxd.UUCP> <634@pucc-k>, <18087@arizona.UUCP>
Organization: U of Waterloo, Ontario
Lines: 220

Extracts from VMSHARE, a community of members of Share who are active
and voting members of the VM Committee.  Names have been deleted where
appropriate.

----------------------------------------------------------------
Comment in "Yes, It Runs on Mainframes" by Donald O'Shea, Amdahl
Director of UTS products in ComputerWorld Extra, Sept. 26, 1984:  "We
believe, in  fact, that there will be only two important operating
systems by 1990 - MVS and Unix."

Reply in VMSHARE: "That's like a choice between death by hanging, or
death by firing squad."
-------------------------------------------------------------
...folks who have used real DEC UNIX don't like it (maybe the same
way we won't like MVS CMS).
-------------------------------------------------------------
Yes we actually have UTS.  We have been running it since last August,
and are very pleased with the work done by Amdahl.  UTS is Unix version
7, with improvements such as a 3270 editor.  We also run Unix V6 with
PWB on a PDP 11/70, Berkeley Unix on a VAX, Bell's System 3 on a couple
of Plexus's (Plexi?).  Anyway, all our Unix's are more or less
incompatible with each other (so much for 'portable' opereating
systems).

...Incidentally, many of our UTS users cannot stand DEC Unix.
They often think that VT-100s are very strange terminals compared with
3270s.  Other comparisons are interesting;  about one in three think
Unix is much better than CMS, one in three think CMS is much better
than Unix, and the rest think that the two systems are almost
identical!
---------------------------------------------------------------
[Running UTS,] the IBM architecture easily outperforms the competition
too; even with VM, a 4341 MG2 is faster than the VAX/780, and response
times are much better (mainly 'cos the DEC screens are so slow).
---------------------------------------------------------------
There are plenty of things wrong with Unix, but they do not seem
inherent in the design.  With software suppliers really having to
compete with each other, it should be posible to develop it into an
excellent operating system.
---------------------------------------------------------------
I can see many advantages in UNIX(*) from the point of view of the
professional programmer, but I'm not sure it could hack it in the
general user environment we're supporting with VM.  Has anybody thought
about a serious comparison?  I think everyone might be interested in
the results.
--------------------------------------------------------------
A great many of our users who use both CMS and UNIX express fervant
preference for UNIX.  None of these people are professional pro-
grammers.
--------------------------------------------------------------
I would be the first to admit that UNIX is far from a perfect system.
It does have many advantages over CMS, however, for both the
professional programmer as well as the novice user.  Here are some of
my favorites:

1. The File System.  There is no question that UNIX achieves its
   greatest degree of user-friendliness through its use of a single
   integrated file system.  CMS suffers greatly through two problems in
   its file system:  First, the inability to share data in a natural
   and convenient manner at a file level, and second in its flat file
   structure.  Have you noticed that once a minidisk swells to more
   than 300 or so files that it becomes unmanageable?  The CMS
   filename-filetype convention is nice, but when I have 150 SCRIPT
   files on my disk the utility of the filetype as a structuring aid is
   called into question.  By the use of directories, UNIX enables me to
   structure the files under my ownership in such a manner that ls (the
   UNIX equivalent to LISTFILE) never has to display more than a
   screen's worth of data.  Humanly this is much more manageable than
   having to stare at pages of LISTFILE output (even if it is displayed
   fullscreen via FILELIST).

2. Modularity.  In UNIX, the fundamental programming concept is the
   tool.  How many times have you cursed (albeit silently) under your
   breath at a CMS command for not supporting a STACK option which
   would enable you to use the output of one program as the input of
   another?  If you're into CMS command writing, how many times have
   you cursed at having to re-write that same STACK code in each and
   every one of your commands?  One of the big differences between CMS
   and UNIX is that CMS assumes that the output of a program is
   destined to be read (or printed) while UNIX assumes that the
   consumer of the output of one program is most likely to be another
   program.  This results in a (some would say too) large collection of
   primitive functions which enable the experience, and even the
   not-so-experienced UNIX user to accomplish a task by combining
   programs in novel ways rather than continually writing new
   programs.

3. Control.  How much discussion has there been in recent years
   concerning "padded cell" environments?  UNIX is DESIGNED around the
   concept!  A padded cell can be as open or as closed as its author
   desires (the UNIX command interpreter, the shell, is an "open" cell)
   but there is no such thing as a CP mode to trap the unwary or
   confuse the innocent.
--------------------------------------------------------------------
Things that I do like are
  1. The 'pipe' system for sending output from one command to the next.
     Incidentally I do know of an internal IBM program that does this
     for CMS!
  2. The fact that each program runs in its own process so that
     any program can call any other program (but watch out for
     problems if both try to process the same file!)
  3. The file system. The tree structure certainly makes management
    of a large number of files much easier, but I miss the file types.
     File sharing is easy because of this structure, but read/write access
     is not so easy.
We have both UNIX and CMS, CMS is currently far more popular but UNIX
has not been around all that long yet.
--------------------------------------------------------------------
I think its interesting how level of expertise reflects in our opinion
of the virtues/vices of something.  One of the primary vices of UNIX IS
the weird command language. For example:

     cat          ----   type out a file on your terminal
     rm           ----   delete the file
     grep   ...   ----   search file for a pattern

Admittedly,  the commands DO other things (ie. they are more flexible
than the simple description I've provided) but, in fact, this is their
main use.  Also, I shouldn't have used < and > because thats how the
shell decides where to get the input and place the output for
commands.  Another major problem is the coding of option flags; for
example, I find syntax like:   "xyz somethingorother -x -n -t -pns"
v-e-r-y  frustrating.  HOWEVER, it is the very terseness of the command
and the ability to specify options in a shorthand way that seasoned
UNIX users love.

The biggest complaint about 327x systems most UNIX hacks have is the
inability of the editor, because of the hardware, to deal with each
character as it is input.  This ability is what gives UNIX the power to
support a full screen editor in a text manipulation sense. That is, a
single keystroke can be used to mark, pick up, move, or otherwise
operate on a block of text.  It also gives the ability to support truly
interactive programs such as games (I've seen ROGUE, SPACE INVADERS,
and even PAC-MAN) on UNIX systems. Its impressive that it can be done,
and I can't imagine the tedium of even APPROACHING that level of
interactivity (?) on a 327x.  Its also EXTREMELY expensive to do i/o
that way.  If you look at what a player or two of one of those games
can do to a system (an i/o interrupt per byte, remember) you quickly
realize why you don't let management catch you at it on prime shift.
Furthermore, I really like Xedit, and I don't think it lacks anything
in the power required to do the editing function.  Except for word
processing applications, byte at a time control seems excessive to
me..., certainly unrequired for professional programmers, or even
authors using a formatter (nroff and troff, by the way, insert format
control words just like script, so I don't know what the fancy editor
buys you there either).

I agree with the comment on reliability.  Things seem to break for odd
reasons (my UNIX colleagues will string me up if they read this), and
the machine (the one I'm most familiar with at the moment is a VAX
11/780 with 4Meg of memory -- I think this is about a 158-III) doesn't
seem to be able to stand up well to a large user community.  25 users
or so and things get perceptably slower (although to be fair,
performance is consistently good right up to that point).

As for the padded cell, yes, the shell is the perfect (ultimate) cell.
Its perfection lies in its flexibility.  But how much different is that
from CMS, really, if you view  CMS as a command processor... a shell of
its own.  If you think of CMS as a shell, by the way, its not hard to
extend the analogy that CMS is what TSO should have been, therefore,
run CMS in an MVS address space.

I agree with [XXXXX] that not enough has been said about that other VM
component ...  CP.   When it comes down to it (despite that fact that
CMS managed to turn out pretty well), CMS might have been as bad as
TSO, or as good as UNIX without influencing what is undeniably the best
feature of VM, the virtual machine architecture itself.  In no other
system can I have online users running CICS, ROSCOE (in an OS guest),
CMS, UNIX (in a UTS guest) and provide decent service to them all.  Not
only that, but make them as independant of one another as possible,
where none can perceptably hurt another.  All that, plus provide a
non-intrusive test environment for systems programmers.
--------------------------------------------------------------------------
I have talked with some people that use UNIX, some really like the file
system, in preference to CMS minidisks. Others say they can never find
their files in UNIX because of the directory structure that contains
the files.
--------------------------------------------------------------------
A few points regarding some of the issues raised in the last few
appends:  First, ANY command language you're not used to is going to
seem a bit strange at first.  This is due both to the confusion of the
new as well as a desire to retain one's old command framework.  cat
stands for catenate and does more than perform a 'type' function.
However, one of the advantages of UNIX is that it is trivial to create
whatever synonyms you wish for commands.

...UNIX can be criticized for its often cryptic commands, but you DO
get used to them.  Command options are admittedly inconsisent, but then
again the same can be said for CMS (some commands support a STACK
option, others don't, for example).  I agree that raw-mode (the ability
to field every keystroke from a terminal) is very expensive, but it is
very nice to be able to operate in this manner on occasion (how often
have you had to explain how to use the RESET key on a 327x to an
impatient user?)
----------------------------------------------------------------
Herb Chong...

I'm user-friendly -- I don't byte, I nybble....

UUCP:  {decvax|utzoo|ihnp4|allegra|clyde}!watmath!watdcsu!herbie
CSNET: herbie%watdcsu@waterloo.csnet
ARPA:  herbie%watdcsu%waterloo.csnet@csnet-relay.arpa
NETNORTH, BITNET: herbie@watdcs, herbie@watdcsu
POST:  Department of Computing Services
       University of Waterloo  
       Waterloo, ON
       N2L 3G1 (519)885-1211 x3524
-- 
Herb Chong...

I'm user-friendly -- I don't byte, I nybble....

UUCP:  {decvax|utzoo|ihnp4|allegra|clyde}!watmath!watdcsu!herbie
CSNET: herbie%watdcsu@waterloo.csnet
ARPA:  herbie%watdcsu%waterloo.csnet@csnet-relay.arpa
NETNORTH, BITNET: herbie@watdcs, herbie@watdcsu
POST:  Department of Computing Services
       University of Waterloo  
       Waterloo, ON
       N2L 3G1 (519)885-1211 x3524