Path: utzoo!mnetor!uunet!lll-winken!lll-lcc!lll-tis!ames!hao!gatech!bloom-beacon!bu-cs!bzs
From: bzs@bu-cs.BU.EDU (Barry Shein)
Newsgroups: comp.unix.wizards
Subject: Re: Help us defend against VMS!
Message-ID: <20268@bu-cs.BU.EDU>
Date: 29 Feb 88 17:09:42 GMT
References: <1636@tulum.UUCP>
Organization: Boston U. Comp. Sci.
Lines: 215
In-reply-to: hirai@swatsun.uucp's message of 29 Feb 88 03:53:44 GMT



>Hello Unix Wizards!
>
>	Our campus is almost on the verge of being turned into a VMS
>filled campus due to the lack of knowledge of the person in charage of
>computing services here.  The next couple of months will determine
>what the campus computer scene will be like during the next decade.
>This person has in mind buying Vaxes with VMS, and DECnet with lots of
>money...

The basic problem with VMS is that it locks you into a single hardware
architecture and vendor. In this day and age that severely limits what
you can purchase in computing power. Vaxes vary widely in price but
not much in processing power. For example, a small uVax-II sells for
around $30K and offers a little less than 1 MIP. An 8750 sells for
perhaps $500K and offers a few MIPs. In the near future this range
closes even further, the uVax-3 being around 1/2 the processing power
of the top end with a price range of ten-fold. It's hard to buy worse
price-performance.

The Unix market ranges from PC based Unix systems which the average
student can afford (and this area is expanding rapidly) to the Cray-2,
a premiere super-computer, and just about everything in between. In
the middle market (typical small-medium scale time-sharing) one can
buy Unix systems from various vendors with upwards of ten times the
price performance of VMS.

Unix systems are relatively bundled, beyond mere hardware
considerations most Unix systems right out of the box are completely
useable. It can be supplemented in many significant ways with free or
nearly free (eg.  ~$100 for an entire campus) software. VMS is heavily
unbundled, from day one if you want so much as a compiler you begin
layering heavy costs. And you'll pay a separate price for acquiring
and maintaining software on every CPU running VMS on campus. This will
quickly lock you out of the workstation market, having to add $100K in
basic software costs to 40 VMS workstations can put a real damper on a
typical University's plans, no matter how good the intentions.

Unix is the premiere system for compute intensive areas, such as the
sciences using Fortran. The reason is the vast range of power a
program written to run under Unix presents. As I said, a program
developed on a small, affordable PC or workstation can be copied and
re-run on huge compute engines. Although a lot of the sciences in the
past used VMS they now generally realize that this was an error and
the communities are rapidly switching to Unix, any argument that
science is done on VMS is a false argument of the past. You should
poll major science depts and research labs. If nothing else, the fact
that the Cray and other super-computers run Unix has pushed the
equation in this favor, a person using VMS is essentially locked out
of the entire NSF super-computer initiative. Decnet would tend to
reaffirm this retardation (TCP can be had on VMS but it's sort of like
teaching a pig to dance, speak to VMS sites and they'll tell you what
a general pain in the ass it is to deal with third party vendors,
network software breaking on each O/S release etc etc.)

The typical claim by the campus administrator is to point at all the
myriad applications and big-name software that runs on VMS and doesn't
run on Unix. In the first place, most of this now does run on Unix so
that tends to be an anachronistic view. Another point is that such
admins usually have DP-envy. No one on the campus has any need for any
of the big-name applications the person is bragging about, you're
running an academic environment, not a bank! Get a list of these
applications and you'll see how ludicrous this consideration is.  Unix
tends to vastly dominate the academic community in software
availability.  These admins will sneer at things like X-windows, lisp
(which doesn't cost $10K/node), AI systems etc in preference to their
big-name commercial databases and spread-sheets as if what students
and professors do is not to be taken seriously, then why are they on a
campus? It's no accident that both Athena and Andrew have chosen Unix
for their massive campus computing projects.

Unix makes available the forefront of hardware technology, parallel
processing from companies like Encore, Sequent, BBN (Butterfly), RISC
and the so-called "super-workstations" like Sun/4, MIPS, HP etc. which
can deliver nearly 1MIP/$1K of price. The parallel processor provide
time-sharing systems extending into the hundreds of MIPs, again for
about $1K/MIP. Anything new and innovative runs Unix, not VMS, it
would be foolish to lock an entire academic community out of all this.
I can understand why a bank is not particularly concerned, but why a
(supposedly) active research community? Why lock yourself out of all
this. The VMS salesthings will claim that they're going to do all this
*in the future*, they've been saying that for years and years, and
when they do come out with something it tends to be too little too
late, in name only, like a dual processor 8800 which barely exploits
what tiny parallelism it has.

VMS itself is not an interesting operating system to learn or study.
It is basically a re-work of RSX, an ancient real-time operating
system from the PDP11 (Unix also ran on the PDP11 years ago, but it
has grown in modern ways, as opposed to VMS's habit of just accreting
whatever features were needed to meet the next big govt contract.)

The claim that Unix is somehow less secure than VMS is a red herring.
Unix offers sufficient security for campus systems, you're not the NSA
(again the tactic of arguing that VMS is better for things you don't
need.) More importantly, many Unix systems are available with full
sources for a modest price, typically $1000/campus (it's simply a
matter of your vendor choices, more than you can say for VMS where
there is no choice.) Without the sources you are, at best, at the
mercy of the vendor for security. A huge security hole which is
bringing you to your knees (which happens regularly on VMS, and the
news travels the networks like wildfire) leaves you helpless and at
the vendor's whims as to whether or not they feel like closing the
hole this week, or next month, or put it off for next release.

In fact their concern with only commercial DP makes them *less*
interested in your security problems. Banks don't have malicious
students exploiting security holes and don't tend to notice such
things or complain about them. With Unix and the sources you can at
least plug up the hole by a code change and then call the vendor and
wait for the real fix, at least you'll be up and running until then.
Don't believe that VMS sources are available, it's a lie, demand to
see prices for all items needed such as Decnet sources. Demand to be
told what resources it would take to even manage such sources.  Last I
checked it required the dedication of a few hundred thousand dollars
in hardware (basically, an entire larger Vax with large disks) to
manage sources.

Obviously the sources will also be of enormous benefit in answering
user questions, such as tracking down example code using particular
system calls. You can sort of do this with VMS's microfiche, if you
consider searching through microfiche for a particular system call
usage a good way to spend your time. You can't grep microfiche. Even
then you'll usually find that the way the system application
accomplishes what the user seems to want to do is by exploiting some
privilege you won't want to give to a user (I'm not sure I want to go
into the whole mess of the zillions of VMS "privilege" bits which
you'll never fully understand the implications of and will almost
surely end up giving away the store because some reasonable thing can
only be accomplished by giving a user some dangerous privilege bit,
Unix's single privilege scheme [root or not root] is much more secure,
you just don't give out root privs and you know exactly what can and
cannot be done by the two sets of users on your system, who wants to
calculate the permutations of 30+ priv bits and what they might imply
singly and in combination?.)

The programming and system interfaces in VMS are arcane and just a
hodgepodge of features, there's no particular underlying design
philosophy, just whatever marketing wanted this week. Although VMS has
some interesting software features it's nearly impossible for anyone
but a very experienced programmer to take advantage of these.  This is
not really a damnation of VMS, VMS is a platform for delivering
turnkey applications software, like databases in commercial
environments for people who wouldn't think of programming in general,
just data entry and report generation. I'm *sure* this is
representative of your needs (hah!)

In an academic community one merely has to go into a campus bookstore
to see another argument. Look at all the Unix books! Where are the VMS
books?

There are none. A complete set of Unix manuals costs less than $100, a
more than sufficient set costs perhaps $50. A complete set of VMS docs
costs several hundred dollars, no student or even faculty member
(except the few richest) can afford to own a documentation set for
VMS. There's some on-line help in VMS but it's designed to sell
manuals or supplement them, the details are always missing
(purposely.)

Most Unix systems come with on-line, complete manual sets with the
exact same text used to produce the printed manuals.  Thus, what's the
cost to a student for Unix manuals? For $0 (zero) they can get
everything, if they like manuals in their laps they can buy those for
the cost of a couple of textbooks. To supplement that they can buy any
of dozens of titles on Unix ranging from the structure of the
operating system, systems programming, compiler construction,
applications programming, AI, many programming languages, shell
programming, text processing etc etc. For VMS you'll be lucky to find
two titles (I can only think of one, the Internals book, and that's
hardly a text, oh yeah, there's an assembler textbook, both of those
are about five years old and don't even refer to the current VMS
system so you won't be able to use their code etc.)

So, running courses on VMS will mean foresaking textbooks. Very clever!
Good plan for running an academic environment! It's no accident, the
DEC/VMS crowd has no interest in academia, your sysadmin has DP-envy.

Decnet nearly completely locks you out of wide-area networking, such
as the arpanet. One need only look at the arpanet's University rolls
to see who you are abandoning, merely the foremost schools and
research labs in the country. About 95% of them use Unix systems to
hook up to the arpanet. Decnet is completely useless in this regard.
There are a couple of strange, semi-wide area networks based on DECNET
(few people could name them.)  Perhaps one or two of your faculty
would like to be on them. You should buy them a microvax and get on
with the rest of the campus' needs, don't let the tail wag the dog.

And you can forget uucp and usenet entirely, which means no e-mail
to vendors etc.

In summary, buying into VMS for a campus is buying into the past in a
pathetic, nearly necrophilic way for an academic community. It locks
them out of the mainstream in Computer Science, Engineering, the
Sciences and many of the humanities (all the multi-media projects of
any interest are being done on either Unix or or Macintosh/PC
systems.) It has very little to offer an academic community for either
research or coursework. It is flying in the face of nearly all trends
in computing today and doing so at such a high dollar price that it
borders on irresponsible. This is not to say that there is no need for
even one VMS system on your campus, there probably is. But using it as
a campus standard is irresponsible and completely without merit or
rational justification and will cripple academic computing for years
to come. What other campuses do this?

This is not a religious flame, I have presented myriad factual basis
for my arguments. VMS people like to claim religious flame and
"chocolate vs vanilla!" arguments. This is because they cannot deal
with the real issues so making it a political war can only act to
their advantage. Avoid the issues, get the opponents fired, scare a
campus administrator with false promises of donations etc.

Unfortunately you may be up against an insidious cancer you only
barely understand which will manipulate your organization in ways you
will regret.

	-Barry Shein, Boston University