Path: utzoo!mnetor!uunet!husc6!bloom-beacon!gatech!mcnc!decvax!decwrl!hplabs!hp-pcd!uoregon!dboyes
From: dboyes@uoregon.UUCP (David Boyes)
Newsgroups: comp.os.vms
Subject: Re: Unix/VMS "wars" & Machine MIPS "ratings"
Message-ID: <1728@uoregon.UUCP>
Date: 22 Mar 88 03:00:11 GMT
References: <36568QAA@PSUVM>
Reply-To: dboyes@drizzle.UUCP (David Boyes)
Organization: University of Oregon, Computer Science, Eugene OR
Lines: 85

In article <36568QAA@PSUVM> QAA@PSUVM.BITNET writes:
>Rather than approach specific things which make one system better than or
>worse than the other, I pose one simple question - What is the definition
>of "compatability" that's used in reference to UNIX?

Compatibility:

The same source program will compile *without modification* on any
machine running the same version of Unix (or a version of Unix that
provides compatible library routines).

>Surely you don't expect a VAX-UNIX program to run on a different UNIX box
>without recompiling it, then where's the compatability between UNIX systems?

Why not? I use the same sources on a Sys Vr3 system and a 4.3bsd
system. Same program runs fine on my beloved IBM 4341. VMS is the only
system I've had to change my ray tracing program for. 

>another may be easier than porting from machines with different operating
>systems, the fact remains, some conversion would most probably need to be
>performed.

Not exactly true. A program that uses the standard library routines
and uses curses for full screen i/o will work without modification on
99.978% of Unix systems. Programs that take advantage of features like
Sys V shared memory or semaphores or V-system task groups will
naturally need rewriting.

>** flame on **
>To all UNIX lovers - give me ONE good reason why "rm" is better than "delete"
>to delete a file.  And of course, there's the "wonderful" "help" on UNIX
>systems - I won't even begin to touch that one...:)
>** flame off **

Explain it as 'rm' *removes* a file from the disk. Inexperienced users
here haven't had too much trouble with that. 

Quite frankly, VMS help is not all that impressive. Any help system
that won't look at help files in my OWN directory first is a bit silly --
how am I supposed to know if it's going to look right without seeing
how it's going to look from the HELP environment? Come on, guys --
even IBM figured this one out. Library files - bah.


>narrow minded.  Sure, I wouldn't argue the fact that there are a lot
>of high powered workstations out there that in raw compute power will
>run circles around your average VAX, but has anyone ever considered
>adding in I/O and other factors into those ratings?  Let's take the
>VAX 8800 as compared to an equivalent 6-12 MIPS workstation.  Can you
>see 300-400 people logged onto the workstation - ALL at the same time?

You're missing the point. The whole idea of workstation computing is
to AVOID 400 people sharing the same CPU. In a distributed
environment, it doesn't make a bit of difference whether I'm running
my 1024x1024 raytracing jobs or a 1Mx1M image enhancement job while
you're editing away on your latest opus. You have your CPU to use in
any way you want and I have my CPU to use in any way I see fit. Let
the big machines do things like handle disk service or sharing
expensive high-speed peripherals like high-volume laser printers, etc.
With inexpensive high-speed network media, it is now really more cost
effective to get a load of workstations and network the heck out of
them using something like NFS. Most of the real use (MY OPINION!!) of
large machines is to access large volumes of shared data -- a task
easily done using NFS and workstations. If I choose to do ray tracing
of an image based on how many people in Illnois voted for Bozo the
Clown, J. Random Luser shouldn't have to care. My work should have no
impact on his session -- and it doesn't, in a workstation environment.

>My point is simple - when you consider the 'power' of a machine - don't
>forget to factor in things such as throughput - unless you're going to
>dedicate the machine to ONE person. 

Why not? With Suns/Apollos/Masscomps/Convexes/etc. being far cheaper
than their approximate DEC equivalents, why should I have to share?
Why shouldn't I throw a job on the machine that will sit and munge for
a week? I can't do that with a large DEC box.

>-Tim Bieling, system coordinator
>Email:  BIELING@PSUARCH.BITNET 

-- 
David Boyes         | ARPA: 556%OREGON1.BITNET@CUNYVM.CUNY.EDU
Systems Division    | BITNET: 556@OREGON1
UO Computing Center | UUCP: dboyes@uoregon.UUCP
'How long d'ya think it'll be before just us oldtimers remember WISCVM?'