Hardware/software schizophrenia Date: Tue, 19 Dec 89 14:08 CST From: uunet!MCC.COM!ai.gooch (William D. Gooch)

[Gyro@REASONING.COM]

	I beg to differ (about its being depressing).  I think if they had
	thought of themselves as a hardware company to begin with,

    To my knowledge, Symbolics has always viewed itself as essentially and
    primarily a hardware company.

Up to a couple of years ago, not only did the people I talked to in
Symbolics tell me that they thought of themselves as essentially a
software company, but they behaved as if they thought of themselves
thus, in terms of their price structure, their relative nonchalance
concerning the issue of CPU performance, and other ways.

Seems to me that if they thought of themselves as primarily and
essentially a hardware company, there wouldn't be a ten thousand
dollar license transfer fee.

	    they would
	have built Ivory and the various products based on it a good three or
	four years earlier, and quite plausibly saved their corporate a*s.

    "three or four years earlier" would have been less than two years after
    they introduced the 3600.  I'm not sure that's a reasonable expectation.

Perhaps I exaggerated.  Two years, anyhow.  -- I don't know what's
"reasonable" in the VLSI world, anyway.  I do know they made an
explicit decision in c. 1983 not to energetically pursue the VLSI
implementation, a decision which has more recently been publicly
repudiated by the original president of Symbolics himself (whose name
escapes me at the moment, alas).

    Keep in mind they introduced one other distinct architecture (the "G"
    machines) in between the 3600 and the Ivory.  In the meantime, no other
    lisp workstation vendor has ever been able to come up with a single new
    architecture, just modified copies of existing ones.  (In hindsight, I
    must grant that the G machine effort, while it was a tactical success,
    may have been a strategic mistake since it delayed the Ivory effort
    somewhat.  However, it certainly didn't look that way at the time.)

Uh -- it sure looked that way to me.  I remember when I first heard
about the G-machines.  I was very surprised and disappointed.

	    If
	there had never been a time when Suns had a better price/performance
	ratio for running Lisp, they would have had a market that might not
	have been very large, but would have been steady and loyal.  As it
	was, the appearance of the Sun 4 at AAAI-87 was the all-but-fatal
	blow.

    I don't think this is a particularly accurate portrayal of the causes of
    Symbolics' difficulties, which are closely tied to the problems in the
    lisp hardware market in general.  It was and is a much more complex
    situation than you suggest. 

I believe I am aware of the complexities, and I stand by my claim.

	I can't resist adding that had they concentrated on hardware volume,
	they would have attracted more third-party software vendors who
	collectively could provide far more functionality than Symbolics ever
	could hack up.

    I'm not sure it would have worked out that way, and I think the feeling
    has always been that Symbolics, because of their size and the uniqueness
    of their hardware, would have been hard pressed to compete pricewise in
    a high-volume market.  They also would have had a lot of trouble ramping
    up their manufacturing facilities fast enough.  As it turned out, their
    sales kept pace with their (high) expectations for the first few years
    after the 3600 was introduced, and they quickly developed a strong
    customer base where there hadn't previously been much of a market at
    all.

Granted, they had trouble with the manufacturing as it was, but such
expertise can be bought.

They didn't notice that facilities doing fancy graphics and AI
research labs were both limited markets, and they didn't try to
cultivate any others.  I believe that had they kept up on performance,
they would at least have kept those two, but they don't seem to have
done that either.

	  They also held their own against the industry-wide slump far
    longer than did most computer vendors.  Unfortunately, as Barmar has
    pointed out, they got a bit overoptimistic about their future prospects
    and grew too fast, then the general slump really hit, and also Lisp
    started becoming a bad word. 

Okay, okay, this I'll grant you: AI itself became a bad word.  That
certainly was a big problem for them.

    I don't quite grasp what all this second-guessing and sniping at
    Symbolics is based on.

The funny thing is, this all started when I pointed out that I didn't
think it was a bad thing that they were finally deciding to be a
hardware company.  I wish them luck; I really do.  I have a completely
irrational emotional attachment to my 3620.  But I'm not holding my
breath for them either.

			    People seem to be unaware that, unlike every
    other vendor that has entered the lisp workstation market (with the
    possible exception of Xerox), Symbolics has profited from it.  Only in a
    few of their very worst quarters did they ever fail to be consistently
    profitable.  Contrast this with the history of Sun, or almost any other
    hardware vendor

Uh, excuse me, profitability is certainly an important thing, but a
profitable shrinking company cannot be considered as successful as an
unprofitable but rapidly growing one.

-- Scott