On Fri, 26 Apr 1996 14:58:08 GMT mark landin said:
>In article <[log in to unmask]>, [log in to unmask] says...
>>
>I have both 9000's and
>>3000's.  Love them both, but we're moving away from the 3000.
>
>How come?
 
Interesting question that I'm beginning to think (in spite of Proposition
3000) my positioning of the 3000 may be unique.  Unix has years of public,
open development, freely available software that makes the CSL look puny,
and way ahead in the networking arena, not to mention generic system support
such as the curses terminal I/O library (check out their supported term
types) and very 'liberal' development environment.  Some of this is good,
but Unix also suffers from a primitive, cryptic, counter-intuitive user
interface at the base level (shells, utilities, etc) and a brute-force
distribution of privileges (priviledge?  I hate that word :-) ) to the
distinction of user versus God with no middle ground.  Unix is a powerful
tool in the proper hands -- countless Ph.D theses have been generated in
the expansion of base Unix tools (check references for regular expressions
in grep, or even deeper, pattern matching in Perl).  The problem is that
this is incomprehensible gibberish to a non-Unix audience; the closest
analogy I can think of is a COBOL programmer looking at RPG source code,
except worse.
 
As someone stated earlier (Wirt?), Unix rules because "everybody's using
it".  Sure they are, Linux is free, you can get Sun Solaris for Intel PCs,
and Unix vendors are waging a price war to land a market share.  But it is
plug-and-play nonsense fighting for the hardware market.  Let's face it,
if you're committed to a Unix platform, you send out bids and anything
goes; few vendors have "added value" of significance to offer to justify
choosing them over low bid.  So you choose your "Unix box"...
 
I wonder how many Unix sales are to NEW Unix customers (I suspect a large
percentage since this factors out the long-term educational and research
areas where Unix has always had a niche).  It's strength is in the wide
diversity of applications and utilities, many free (Internet legacy), and
if you have a Unix-guru on staff to hide the management/operational
overhead, you might get away with it for awhile.  But the reality of Unix
eventually becomes apparent... I just finished a grad course in Internet
programming (HTML, perl, JAVA, and JavaScript) and the Unix components
have an incredibly high learning curve.  Once you gain some literacy of
the language you can do incredible things; but the learning process is
incredibly slow and there is no "manual" to learn from.  Unix has "man"
pages, yes, but they are intended for reference use and offer no practical
examples of use.  The MPE Posix User's Guide (not exact title, I don't
have it handy) has some excellent material in it for practical use; but
it isn't on the LaserROM nor widely distributed in paper format.
 
Which brings me to my "unique" point (perhaps) that if the 3000 can run
the Unix apps, we're really getting somewhere.  Unix looses the "everyone
else uses it" edge and the 3000 footholds take effect.  Unfortunately the
hot Unix apps require Oracle, Sybase, or some other RDBMS that costs an
arm and a leg on the 3000 and always lags behind the competition in its
feature set and availability.  Still...
 
After some practical Unix exposure (possibly after some expensive HP
training classes, or whatever vendor you choose) you have to eventually
face the pitfalls and complexity of Unix management and operations.  As
Stan noted, not only do you need more CPU, memory, and disk storage; you
also need systems personnel dedicated to the welfare of the beast.  Buy
the O'Reiley (sp?) series of Unix books and try to align to the mindset.
 
There was an April Fool's day joke a year or two ago where Ritchie and
others confessed to the Unix "hoax" - Multics was MIT's groundbreaking
timeshare system, and Unix was the counterpart for the workstation as a
subtle joke (Unix vs Multics, One [user] versus many] and some C jokes.
Perhaps not that far from the truth.  It runs on a PC, and now even the
PowerPC, it's not rocket science from implementation.  It is extremely
flexible in the APIs at the cost of security, segmentation faults,
fschecks, and so forth.  It has strengths in the short term (cheap, many
free packages, universal acceptance, etc) but will it survive the long
term?  As Wirt mentioned, it may die out as a fad as he suggests, or
else collapse on it's own shortcomings (my prediction).  Only NT is a
ground-up rebuild of an OS with Unix in mind in ages, but it lacks much
of the appealing software (so far, but they're growing) of Unix.  MPE
took a bold step forward with Posix, but haven't followed through with
the necessary add-ons to make it truly and "open" system.  "Open system"
is such a buzzword.  If MPE and HP-UX are both "Open" why are there two
software divisions?  Are there any two platforms that are 100% source
independent?  (maybe there are, but point remains "some" porting effort
is inherent in most cross-platform apps).
 
One noteful point with regard to Posix/Open Systems/etc., in various
surveys of the MPE user base, few people have bothered to explore this
new environment.  Many of those that have tried were buried in the esoteric
quirks and shortcomings of the initial offering.  Some of those who stayed
would like to see some things changed (is Posix a dead product now?) to
make it really valuable.
 
MPE users, in the traditional sense, embrace their platform for it's clear
and obvious strengths in the face of popularity.  This has never changed.
MPE nor the 3000 has ever clearly "dominated" the hardware or software
market as the principal player; but they have had their impact as measured
by the Datapro (Datamation?) survey as ranking #1 in overall customer
satisfaction.  That speaks volumes.  It works well, and rarely, if ever
ruins your day.  MPE has a solid foundation that should never be lost; but
to face reality, the applications arena is suffering because vendors have
shifted to Unix platforms where the numbers are more favorable.  As the MPE
market shrinks, 3rd party offerings will decrease in numbers and increase
in price to recover porting efforts.  Posix brought us to the peak of the
summit, but still fell short (IMHO we're that close; others may disagree).
If we can't crest the hill to compete in the current market, we may well
follow RTE into the abyss of obscurity/obsolescence.
 
We have NCSA httpd (despite it's faults, but then look at OpenMarket which
had some serious effort put into it), Gnu tools, a BSD library, sendmail,
lynx, and more recently perl5, freevt3k, etc.  Some argue Posix is a cryptic
and confusing add-on to MPE, yet considering your alternative is Unix, you
should count your blessings.  Once you install a Posix app, you can run most
of them from an MPE session/job, have journalled file system, integrated
backups, and other advantages without day-to-day dealings with a shell.
 
All this and MPE too!  I am not suggesting MPE "turns into" Unix.  There is
so much that *could* happen, but few MPE users are interested in considering
let alone using these new tools.  Most upsetting are those users who are now
facing a mandated Unix conversion that still scoff at Posix.  With some
dilligence you might be surprised what you can do with your 3000.  When the
line is boldly drawn between Unix and MPE, we're a struggling piece of
history.  Keep that line fuzzy and we can compete in the applications arena.
I don't think anyone has left MPE by choice, but rather by demands of their
outside applications.  I know this first-hand, we're facing such a choice
and perl5 was one such pre-requisite (thus my interest in that port).
 
In summary, I appreciate the passion with which Wirt and others defend MPE
for MPE's sake; there is no argument if you have a stable environment and
no major developments in progress.  I understand the lure of Unix, but I
want no part of it -- if you type 'exit' in the Unix shell you don't get a
colon prompt to escape to reality; you are stuck in your "chosen" life.
 
MPE is wonderful; it's low-level structure is immaculate.  Unix is open,
but a widely hacked kernel never designed for speed/ease not reliability.
MPE is a two-bedroom flat on a foundation of solid rock, Unix is a high-rise
condo built on a foundation of sand with discount rates for new tenants.
Carry that a step further and the MPE "flat" has no cable, touch-tone phone,
and only recently offered locations in mixed-housing neighborhoods.
 
Despite opposing views from many of my well-respected readers, Posix is our
last hope for longevity.  Die-hard, mission critical applications and their
customers can only last so long.
 
Respectfully,
Jeff Kell <[log in to unmask]>