Dan Hollis writes:
 
==================================
 
You really don't know a *thing* about standards until you can write a
program that with only minor changes, will compile and run on 15+ completely
different architectures, with very little effort.
 
Case in point: Pixar's Renderman software, originally written to run on
Crays and other super-high-end hardware, took only an afternoon to port to a
486 running Linux. Of course, it runs much slower, but... :-)
 
>If you're dominant in a marketplace, it is simply foolish
>to allow everyone else to play on "a level
>playing field").
 
This statement is totally naive. Without standards there is no
interoperability. Period.
 
Without standards there are no floppy disks, no DATs, no SCSI and no IDE.
Without standards there are no modems. (V.34 etc.)
Without standards there is no internet. (TCP/IP)
Without standards there is no ASCII.
Without standards there are no computers. At least, none worth using.
 
===================================
 
Actually, the statement is not as naive as you make it out to be.
 
Each manufacturer adopts only those standards that it believes it must to
compete in the marketplace in the manner necessary to attract whatever
customer base it is seeking. It is that way now. It has been that way since
the beginning.
 
Part of the problem is that I'm getting old -- and I've heard all of this
before. I began working for RCA Service in 1959, when the digital computer
was only 14 years old. And I remember punched tape and punched cards very
well, when the codes were Baudot, Hollerith, Remington Rand, ASCII and
EBCDIC. Indeed, for the first ten years of my professional life, I used
primarily only Baudot 5-level code for programming *analog* computers and
transmitting their resultant data on a world-wide web of radiotransmitters
using frequency shift keying for 2 to 6000-mile skips (of course, all
transmissions were manually relayed from station to station using paper tape
recordings and often had to be repeated until a good copy was received; the
error check routine used was: three transmissions of the same data, where two
were forward transmissions and one backwards, using the advanced
error-checking algorithm of "thrice-told tales are true." If the simple
digital summarizing verifier found that all three repetitions agreed, the
message was acknowledged and manually relayed onward. But other than the
manual, tedious attribute of it all, the WWW is deja vu all over again).
 
ASCII code existed in the early 1960's as an internal Bell standard, but it
wasn't called ASCII until 1967 when it was chosen by the USA Standards
Institute to become the standard method of communication between any form of
binary communication device. Computer manufacturers (other than IBM) quickly
adopted the new standard, but computers weren't the reason for ASCII's
initial creation, any more than RS-232 or RS-244 were originally designed for
computers. And IBM held onto EBCDIC well into the early 1980's, in defiance
of ASCII, simply because they thought that they dominated the computer
industry (which, by anyone's account, they did).
 
The same is just as true of DAT tapes. DAT stands for digital audio tape, but
it never flourished in the US in that incarnation simply because of the
record industry's great (and reasonable) fear of unbridled piracy. Using DAT
tapes as a computer storage mechanism was an afterthought -- and that
afterthought was almost certainly due to the fact that so many very
inventive, but very cash-strapped, university people (primarily radio
astronomers) had been for years before converting cheap, off-the-shelf VHS
video recorders into high-density digital recorders (a DAT drive is just a
miniaturized VCR).
 
V.34 is a more recent example of the same phenomenon. V.34 was a proprietary
design promulgated by MultiTech, first by gathering other commercial
adherents, and then standardizing it by proposing it to the ITU in Geneva
once enough mass had been gathered. Rockwell had earlier devised another,
incompatible 28.8K standard called V.FC and were first to market with
high-reliability 28.8K modems. For a bit of time (up to about 18 months ago),
Rockwell's 28.8K modems dominated the market. Now most modems are either V.34
or V.34/V.FC, which is to MultiTech's great advantage and to Rockwell's
dismay.
 
Similarly, in the mid-1950's, the NTSC (National Television Standards
Committee) color television standard that was originally devised by RCA
Laboratories won certification by the FCC over Columbia (CBS) Labs
incompatible (and non-backward compatible) transmission protocol (NTSC was an
enormously clever way to transmit color picture information in between the
frequency-sampled "luminance" information of the existing black-and-white US
televison standard).
 
But my favorite example is, of course, HP-IB. In a 1968 "Electronics"
magazine article, HP publicly acknowledged that I and Joe Ellington were the
impetus for creation of the then brand-new HP-IB bus interface. In 1964-5,
Joe and I had taken HP9100 calculators and converted the printer port into a
bidirectional port for the transmission of data from calculator to calculator
and to insect environmental control chambers. HP engineers from Greeley came
to visit us twice to see what we had done because of the enthusiasm of the
local HP office. HP changed the protocol of HP-IB a bit from our original
design, but it remained the same basic design. Once HP began manufacturing
calculators in large quantity to control instruments, they petitioned IEEE to
have it declared a standard, which IEEE did after a few years  and called it
IEEE-488 (but what the hell, IEEE even got a committee together and codified
the standards for the Altair and SWTC computer buses in the late 1970's). At
the peak of IEEE-488's acceptance, there were perhaps a hundred
instrumentation manufacturers building controllers and instruments that
talked IEEE-488, but only HP, to my knowledge, ever adopted it as a computer
peripherals communication standard.
 
When HP declared that HP-IB was obsolete a few years ago, I knew then that I
was getting old -- if that fact wasn't already obvious enough to me.
 
But the bottom line is this: standards in computing are no better now than
they were 30 years ago. And all standards are the result of proprietary
products than won out in the commercial market for one reason or another.
Indeed, you could easily argue that standardization is much worse now than
that it ever has been. In 1966, thirty years ago, you could take a FORTRAN
program and run it on 15 different platforms, just as you suggest now, Dan,
and get it too to run with little or no modification.
 
POSIX is not a hallmark zenith attribute of UNIX -- rather it could just as
easily be characterized as a desparate and possibly last-ditch attempt to
bring some of the same level of portability that existed 30 years ago to an
operating system environment that simply got out of control during that
period when the Bell System was prohibited by judicial law from selling or
marketing computer software or equipment.
 
My profound enthusiasm for the HP3000 doesn't stem from it's
intercommunicability abilities. They're about the same as any other computer
(in that whatever's missing will be soon supplied; the trouble is that
machine-to-machine intercommunication standards have always remained a
persistently moving target over the last 30-40 years).  Rather, my enthusiasm
comes from my rather deep belief that the HP3000 -- or some machine very much
like it -- represents the way that all business machines must become in the
very near future. Seventy percent of our customers now run their HP3000s with
no data processing staff on-site. Having a large data processing staff
hovering around any computer is as profound a sign of an overt design failure
as is having several mechanics ride around with you in your car.
Historically, such situations have lasted for only a brief period of time
before the machinery became reliable enough and simple enough to use that it
could be used without having to house gurus or wizards on the premises.
 
Computers have taken a bit longer than most other devices to reach that point
than other technologies have, but the HP3000 is closer than any other
computing platform to achieving that ideal. Moreover, that ideal is achieved
only through a very careful proprietary design integration and optimization,
watched over by very attentive design engineers. It is a slow process -- and
all the money in the world can't buy the time necessary to create a perfected
design in a year or two. Your hesitation, Dan, to use Windows NT for critical
business processes is well justified -- but never underestimate the
commitment of one person who has the will and the resources to get a job
done, most especially when his opposition is a committee.
 
Wirt Atmar