25 October

Timing when to do the next big thing

Why does it take so long for an opportunity to "happen"?

I was just asked why something I was involved with happened to take off so late, when all the elements had been in place for many years - why should it be such a big thing then, when it, by all respects, should have been a big thing earlier?

For a businessman, this is the key reason they hate new things - including technology. For all intents and purposes, something should be obvious and have an immediate effect - but often doesn't. Then it gets done again later, and may then be a phenomenal success. Which can be doubly vexing for the innovator (and the innovator's investor).

Even such market savvy ones as Steve Jobs and Intel Corp have screwed up here big time on this one.

I won and lost on this one with the very same item - BSD Unix on the 386 (386BSD). If you'd like to know how ... read more ...

In the early 80's, I managed operating systems development for a microprocessor vendor doing the first 32 bit virtual memory microprocessor. I had been involved with a variety of UNIX related operating systems at Bell Labs, Berkeley, and various "look alike" systems. The VAX Unix combination was incredibly popular as a mature and widespread platform for software development.

Interestingly, it wasn't the fastest, the least expensive, or the most numerically accurate (pre IEEE floating point with "issues"). But it (and the venerable PDP-11 that preceded it) drove the demand for UNIX systems on just about everything. Including the original IBM PC.

Really this machine was IBM's view of what an Apple II should have been, updated to 10x the memory and twice the word length. But it was still no VAX or even PDP-11. So while you could get a "nix" contorted to appear to run on it, there would be compromises. And these compromises would need to be passed on to any applications that would run as well.

Worse yet, the compromises wouldn't be the same - many groups created "Unix-like" systems with different approaches - none common. So you might need to have dozens of versions of an application to run on all of them, even on the same architecture. They were all too arrogant to agree on a fraction of commonality, let alone a whole. But this did give rise to "Unix standardization" - which took decades. And is still happening.

Worse yet for Intel, the very strategy of reverse compatibility that helped them sell successive evolutions of X86 architecture also limited them. People viewed them identically - e.g. that like with the original IBM PC, you should use a segmented programming model with 16 bit registers on all the versions, carrying the software model forward where it didn't need to be. This psychological barrier grew to be a market barrier for years(!).

I tried to sell a modern form of UNIX (flat 32bit) for more than 2 years before 386BSD began. And even while doing it, the 386 system parked next to the VAX in the Berkeley server room was routinely ridiculed - it was so small, so cheap, so ... insignificant. The project was routinely insulted - when I visited Steve Jobs NeXT, they had low regard for Intel's processors, even though Motorola couldn't keep up. Even Intel itself had little regard for UNIX, thinking it an obsolete system, so that when we'd stopped by for compatibility checking on new processor enhancements, they'd also imagine that "Wintel" was the only option.

Years later I ran into David House of Intel. He told me that he couldn't understand why Unix on Intel, specifically Berkeley UNIX, took so long. I asked him who he contacted at Berkeley about it, and the names he gave me were the ones working on (and payed to consult for) Motorola and DEC (Vax). They never contacted me! Nor did my supposed associates at Berkeley, sitting at my elbows, working and otherwise sharing with ... ever tell me.

What is terrifically ironic about this, is these are the same people who attempted to make money off of selling a X86 BSD - they injured it from the start from self-dealing - trite money - Berkeley has always had a difficulty in dealing with businesses (undisclosed relationships and un-invested), a conflict that Stanford has resolved (disclosed and invested). Grant stealing and blackballing. Berkeley continues to struggle with this.

At the time Linux started to gain steam.

BSD and Linux had very different aims originally. BSD faded away because of an intentionally poor understanding of the market. But that is a story for another time.

Why 386BSD was briefly extremely successful was that it attempted to recapture the same environment of the VAX on the X86 - you could cut between the two deterministically and with little cost. The reasons for this are documented in the "Porting UNIX to the 386" series I did. The point was to lower the barriers to use - then it got used - the missing reason for X86 UNIX success in the first place, which could have been done years before.

So where did it fail? BSD had its purity movement, based on many competing "cults of personality". Linux had one cult of personality - around Linus. Had no interest in the job myself, which infuriated many as much as if I were to have had.

The purity movements moved BSD backwards for a few critical years, doing remarkable damage, and burning many bridges between people who would have to cooperate to succeed - they'd rather kill each other first. Which they did.

Meanwhile even Steve Jobs came around to X86. After almost losing control of X86 to AMD, Intel also finally began to see that control of the architecture had its limits.

Technology doesn't happen in a vacuum - it happens in a social context. If you listen to how people see application of it, you can get a sense of timing and how to factor things in "just so". You can't force a vision of what it will be. Microsoft relearned this with Vista too.

Understand your market for a product at your peril.

Posted by william at 23:31 | Comments (0) | Trackbacks (0)
<< Global Retrenchment - Retreat, Not Surrender | Main | Intel in the 80's is like the US in 2009 >>
Re: Timing when to do the next big thing

I had an AT&T PC with UNIX in the mid 80's. Why did it take until the early '90s for the 386 version?

Posted by: J at October 25,2009 23:45
Re: Timing when to do the next big thing

Dear J

Thank you for your question. I vexed over this issue during those years.

In short, the non-viable contortions to get UNIX on a 8088 were easier to enhance than get rid of on 186, 286 architectures. So when the 386 came out, they used it like a 286 reverse compatibly. Since that was like how "Wintel" did it, that certainly must be the best way - they thought. Doing a native true UNIX port wasn't desired for the virtues it would bring - no one wanted to fund it, so it had to be done for free before becoming accepted as valuable.

Posted by: William at October 25,2009 23:49
There are no trackbacks.
Post a comment