[TUHS] Why Linux not another PC/UNIX [was Mach for i386 ...]
clemc at ccc.com
Thu Feb 23 03:00:59 AEST 2017
On Wed, Feb 22, 2017 at 11:11 AM, Larry McVoy <lm at mcvoy.com> wrote:
> It was not freely available.
> You keep saying that 386BSD was up for FTP as a poorly kept secret.
> Why was it secret? Because it wasn't legal to get it.
Fair enough.. but that was the point where the CRSG folks were beginning
to push back a little. Bostic has already suggested it had been
rewritten to the "CSRG Management" and folks were beginning the rethink the
whole issue at UCB. Remember BSDi does not yet exist. Before BSDi can
come to be, the idea of BSD for being "freely" available (for
any processor) has to be broached.
> A lot of us were pretty sick of that legal bullshit.
Actually - that was exactly the point. A lot of folks did not think that
the AT&T copyright mattered any more. I think most of us at the time we
sick of it. The code base had been rewritten. But of course, none of
us were lawyers. It was pretty clear an awful lot of the code was not
"direct derivative works" of the Research base at this point. /usr/ucb
was >> /usr/bin. The boot system had pretty much been completely redone.
Tools like Sam's config subsystem did not exist in the AT&T code base.
Plus, by now BSD has already switch compilers, so the whole PCC thread is
gone. Even the rest of user level tools had slowly been rewritten. Keith
had been replacing anything in /usr/bin or /bin with code from a difference
provenance. That's part of why he rewrite ex/vi - so they it could not
claimed to have been based on ed anymore.
But you are right .. no one was really sure what was going to happen. So
the BSD/386 iso's were hidden, so at least officially they could say the
only people that we supposed to have access were the BSD licensees. But
as I say it was a well known "secret."
Take someone like myself at time. Officially, I'm a contractor. I had
been working at Masscomp and Stellar before this. I had been a student at
UCB and worked with the CSRG folks. At this point, I had branched out on
my own. I'm whoring myself to anyone that will pay me to hack on UNIX.
At the time I had a gig with NCR. I don't personally own AT&T licenses.
The folks I work for do. NCR is working on what would become SVR4 BTW.
I'm clearly "mentally contaminated" with the AT&T IP. But I know the BSD
code base pretty well also.
Since, I had been a BSD hacker and he was a friend of mine from the old
days, I hear that Bill Jolitz is having issues with the AT/Disk
controller. I offer to help him cause I have access to the WD1003
controller doc and think I know what is going on. Bill sends me the url
for the system so I can down load it mess with it. I use my Wyse 386:16 at
home in MA to load it (and it fails). So, I start hacking, redo the driver
and send it back to Bill. Bill likes my new driver and switches it....
etc... now I have the sources ... too "unofficially." BTW: the truth is
my customer (NCR) is a licensee and I have access to the AT&T code South
Carolina when I am there. I'm very, very careful to never mix code
bases. But if the Judge ever asked, I justify that NCR is license from UCB
who is licensed from AT&T and I'm licenses from NCR. But I will also look
the judge in the eye and insist never, never did the UCB code run on
anything owned by or paid for by NCR.
Could I have been sued, I don't know. I'm not a lawyer, but as I have had
copyright law explained to me, I should have been ok. Trade Secret on the
other hand, clearly, I had seen AT&T's IP so I was contaminated. But I
had been was contaminated at CMU, 15-20 years earlier with the same AT&T IP.
> Linux didn't
> have that problem.
And BSD did not either in practice, although as I said, once the suit was
filed, you are right. A lot us, myself in included got scared. So we
switched... because we thought (incorrectly) that Linux was free of the
legal burden (it wasn't) - we were all contaminated, Linus, Tannenbaum,
you, me, the Russias :-). As I said, the cow as out of the barn and barn
had burn downed long ago, so trying to keep the IP off the market, had
little realistic chance, IMO.
It did not matter if it was called UNIX, Minux, Linux, Idris, Coherent or
Larry and Clem's cool new OS. They all had the problem.
> I think most hackers wanted it to be free, though even there it was
> sort of hit and miss.
I agree. Or really closer to free than $1K. I was willing to pay $50-100.
> The BSDi guys thought they saw a market
> opportunity so they weren't so excited about free.
Yup, and I understood that too. I had done 2 startup's by then, so I
already had a feel for the cost of care and feeding of hackers like you and
me. IIRC, BSDi's "nut" those days was about $2-3M per year, so at $1K a
copy they need a few thousand copies per year to may salaries and expenses.
What they really needed was a larger firm, say some one like ILM or one
of the National Labs that was using the technology for a critical thing and
would pay many K per year for support. But they had not been figure out
They were getting there ... but then the law suite came.
Having had this conversation with him at one point, I actually think, Kolstad
had already gotten to "service model" - but the law suite killed the goose.
They never recovered. So Linux with RH and Cygnus behind them filled
So in summary -- I think the two parts of Christiansen disruption theory we
all agree is that it was
- the PC/386 based UNIX that was the key driver and
- the acquisition cost for the end user had to get low enough to make it
Those other stuff we will never completely know. I suspect it depending
on how to you look at it, personal experience. I acknowledge a number of
your points and can see how you might come to such conclusions. That
said, I do stick by own belief of the root cause, but not coming to
complete agreement is ok with me.
I definitely learned a bit from you and Dan's view and I thank you both.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the TUHS