OS's, GUI's and Life

Ernest Retzel 1535 49118 ernest at lenti.med.umn.edu
Sun Mar 17 20:09:08 EST 1991


I am generally not a "post-er," but the recent operating system discussions 
have finally prodded me into it.  First, I would like to identify myself as a
card-carrying, working virologist.  Second, [indulge me, please], I would like
to make a public apology to Julie Ryals and Dave Kristofferson for some e-mail
conversations in which I naively took them to task for something last fall [no
one else is likely to understand that, but hopefully they will].  In following
the OS discussion recently, I now have a better understanding of the situation.

What pushed me to post was a recent article regarding "easy vs powerful
operating systems," which was well-written and persuasive, but also somewhat
disturbing and misleading; that it was subsequently applauded caused even more
of a twinge.

The first disturbing thing is that it sets up two classes of humanity, the
Biologists, who work for a living, and the Power-User, the Programmer, the
previously-mentioned Nemesis of Civilization, the System Administrator, and,
probably, the forgotten System Analyst.  I am not sure where the latter fit into
life-forms generally, but it appears that all of Them Cannot Be Trusted.  One 
cannot, apparently, cross the proverbial line without becoming Lost Forever.   
8^)

The second thing is that there is some apparent misunderstanding between
operating systems and graphical use interfaces [GUI's].  The discussion mostly
centers around Mac applications [programs] and compares them to UNIX, which is
a bare-bones operating system, OVER WHICH can be layered a GUI.  What is being
compared, then, is not operating systems, but corporate philosophies: Apple
enforces their idea of a standard, Sun and the rest [DEC, etc,etc,etc] allows
the programmer or the user to make their choice.  This thread was, in fact,
initiated by a question/statement from Don Gilbert regarding exactly that
choice [which I would like to see a summary of responses to sometime].  When
you are discussing operating systems, however, it is kind of an
apples-and-oranges discussion; a System Analyst friend of mine once made the
offhand comment, "If it comes on a floppy disk, it is not an operating system."
UNIX is something over 100 MB of disk space on its own.  And, considering that,
one really does not have to know much of it to *use* it.  A few commands, and
one is on their proverbial way.

If one can do everything they want on a personal computer, fine; if one is
doing mostly just word-processing on their computers, they are probably being
underused; if, however, you are "bumping your head" personal computers, then it
may in fact be time for some other solution, and that is most likely a UNIX 
machine of some sort.  As Bill Pearson pointed out, the price-performance of
workstations is really quite dramatic.  The possibilities for using them are
nearly endless.  In our department, we have gone from one workstation to 15,
acquired in ones and twos, *by popular demand.*  Suns now outnumber either Macs
or PC's [but not both, yet], and are replacing personal computers in individual
labs.  Even our secretaries use Suns [and UNIX...], and are, in fact, some of 
the strongest supporters of the system [because it saves them time...].

We are working on a very different paradigm than some of those that are
expressed in previous posts; in those I am referring to, the opinion is that 
computers are something to be used when the Real Work is done, as analysis 
tools; our general opinion is that computational tools can be *part of* an 
experiment, and can, in fact, be used to *give direction* to experimental 
approachs.

We are also truly net-based; locally, we have the Suns I mentioned above and
seven disks totaling about 4 GB of space.  Using UNIX built-ins, like Network
File System [NFS], we can mount disks and use computers anywhere, and do.  And
I do not need to know about them on a personal basis; UNIX knows about
them--that is something it is really good at.  By way of example, the Minnesota
Supercomputer Center has several resources that we use, specifically a Cray XMP
and a Cray2 [512 *megawords* of RAM, in 64 bit words]; both of these are UNIX
machines.  They also have some 200 GB in their disk farm [think infinite...].
With a properly designed program and GUI, I can set up a program on a
workstation here in the department, run the job on the Cray, see the graphical
output on my workstation *as if it had run there,* and never really know, or
care, where or how it happened.

That is not to say that most programs were designed for this, and they need to
be. Most programs do not operate in this mode [I am pleased to say that there 
are a fair group of programs that have been evolving here that do, however].  
Intelligenetics, to their credit, started out this way, with
workstation-based programs, some even making use of Sun Core Graphics.  But we
all wanted to treat the Suns like a VAX, so their programs devolved into TTY
based programs, instead of evolving on the workstation track.  Now, it sounds
as if they are moving that way again, and that is really nice to hear.  We are
once again back at Don Gilbert's question; reality is, however, that building
graphical interfaces costs money [see Chris Dow's comments]; I have been told
that if you start from scratch, and it takes x hours to implement an algorithm,
it might take 5x hours to do the I/O, and 25-50x hours to put a real interface
on it.  With the UNIX market so small, cost recovery is tough for any of the
suppliers of commercial software.  Will we *pay* for that GUI on that program
we want?  Maybe more to the point, will the granting agencies?

I have, by the way, found it very easy to teach molecular biology to computer
scientists, whether they are programmers, grad students, system analysts or
faculty.  Our problems, when *we* understand them, are almost intuitive to
them.  They can and do create some beautiful, usable solutions, sometimes
beyond my own expectations.  Try to hear Paul Bieganski, Dan Davison, Chris Dow,
Brian Fristensky, Don Gilbert, Rob Harper, Dave Kristofferson, Bill Pearson, 
Tom Schneider, Brian Smith, Roy Smith, the NCIFCRF folks, the LANL folks, and 
the NLM-NCBI folks as they try to unfold what we could do; they are educating, 
not converting.

Final aside--this is bionet.*software* -- OS's are software, as are GUI's, 
and I can't imagine a better forum to discuss them in.

E-mail flames to me personally, or, preferably, to /dev/null.

Cheers,
Ernie Retzel
Dept. of Microbiology
University of Minnesota

ernest at lenti.med.umn.edu




More information about the Bio-soft mailing list