[COFF] Re: [ih] Fwd: Some Berkeley Unix history - too many PHDs per packet

[Clem Cole <clemc@ccc.com>]

[-- Attachment #1: Type: text/plain, Size: 10921 bytes --]

below...   [Dropping IH list].

On Sun, Mar 10, 2024 at 6:05 AM steve jenkin <sjenkin@canb.auug.org.au>
wrote:

>
>
> > On 10 Mar 2024, at 06:52, Clem Cole <clemc@ccc.com> wrote:
> >
> > That said, a different license for UNIX-based IP could be granted by the
> Regents of the University of CA and managed by its  'Industrial Laison's
> Office" at UCB (the 'IOL' - the same folks that brought licenses for tools
> like SPICE, SPLICE, MOTIS, et al).
>
I'm not sure if you are catching that the Regents IOL (Industrial
Laison's Office) [part of the UCB EE Department] and the DARPA’s IPTO
(Information Processing Technology Office) - which was originally part of
the US Air Force, then US DOD,* etc*.. [and went through a number of name
changes].   The latter group originally led and managed a small part of the
US Gov DOD projects.    Its history is best spelled out in Katie Hafner's
wonderful book: "*Where Wizards Stay Up Late*" - ISBN 9780684832678. (More
in a minute).

The former, the IOL, managed the external relationships for the EE
Department (and later EECS when they created the CS division of EE).  It
was set up initially in the latter part of the 1960s by my thesis advisor,
the late Donald O Pederson(*a.k.a.*dop) ---  and, as I said, the folks that
brought you SPICE, SPLICE, and the like].   It already had a way to license
and distribute technology from EE to external organizations [using an idea
that would later be called 'open source.'  He was famous for saying, *"Unlike
our friends across the bay or on the east coast, we give everything away.
That way, I get to go in the back door and see what they are doing.   If I
sell our tools, I use the front door like all salesmen."*   The circa 1977
"Berkeley Software Distribution" for UNIX came from the IOL, as did other
distributions they had been managing since about 1967 or so.




>
> In the last while I’ve read about DARPA’s IPTO (Information Processing
> Technology Office) 1962-1986
>  and how they (generously) funded a very diverse range of projects for
> extended durations.
>
> Alan Kay comments that $1M was small beer to DARPA, who were investing
> billions in R&D every year.

Be careful. The US government, via DOD (and DOE), was funding billions,
while DARPA was a small and mostly forgotten backwater the USAF originally
had set up. As I said, see Katie Hafner's book for more details. $1M was a
big deal to DARPA. But compared to funding a new fighter or a new air craft
carrier, DARPA projects were small potatoes.

>
>
> It was a boom time for US computing research - funders with vision, deep
> pockets and patience :)
>
No, the boom was the Cold War and the space race.  That was driving core
tech.    CS Research just hitched its wagon to those engines.  Things like
the ARPANet were funded to solve what the ARMY called the 'radar problem.:
How(during a nuclear strike) are you able to keep disparate command centers
informed and in sync?

>
> I can’t find my source now, nor any list of IPTO’s contracts given to UCB
> ( or given to anyone ).
>
> UCB - Berkeley - got many contracts, time-sharing / SDS-940, Ingres,
> TCP/IP in the Unix kernel and RISC processing.
>
Yikes -- having lived it. I fear you may be confusing and mixing some
things up - certainly order, and what beget what.  First, UCB was very late
to the DARPA world. Note that the first ARPAnet IMP semi-available to UCB
was at LBL (up the hill).    And while the Regents ran LBL, LANL, Los
Almos, and the like (for DOE, mind you, not DOD).  Furthermore by the time
of CSRG, CSRG did not have the contract for IP/TCP for UNIX -- BBN did.  *CSRG
had a contract from DARPA to support the UNIX kernel.* These are the
sources of famous issues and questions WRT created.  The concept of
sockets(2) was a CSRG [Bill Joy ism -- actually to counter Rashid's ports()
idea in Accent].   The IP stack (and support) *was supposed *to be from BBN
(and it originally was -- you can see at least one early BBN distribution
in the TUHS archives.  BTW, Ingres was partially funded by DOD via DARPA
and predates CSRG by about 4 or 5 years.  Fateman got a contract to move
MAXIMA from ITS to UNIX (and create Franz LISP).   This was the origin of
the original kernel work.  Frankly, I don't remember who funded that; but
I'm not sure it was DARPA.  I think it may have been one of the national
labs (DOE) that was using Maxima.  FWIW: the Ingres ARPAnet connection was
a  'very distant host' interface to one of the 4 ports on the LBL IMP.

I'm not sure who funded Patterson during the RISC work.  I know my thesis
was funded by industrial folks as as well as DOE grant, not a DOD one.

I just thought of another interesting factiod.    Mind you, the BSD sources
were free - which I'm sure caused a number of UNIX vendors to stop there
(per dop's genius of going in the back door) since BBN was a commercial
enterprise (and as such was looking for revenue streams ).   The BBN stack
actually cost money for commercial firms.   In the early 1990s, when we
decided to use it, not the UCB code, at Stellar, we had a get a sublicense
for it from BBN.



>
> There was an IPTO director - Bob Taylor or Robert Kahn - that wanted a
> common development platform with IP plus development tools,
> who gave contracts to UCB’s CSRG to do the work.
>
Ouch ... that is not quite right.  Again - get Katie's book.

CSRG is >>much much<< later in the DARPA (or Internet story).  By the time
of CSRG, DARPA had moved inside of DOD a few times.  It was not nearly the
size of the other teams, but it was a real line item. As Alan Kay said, it
was not even noticed when the original work started compared to other DOD
projects.

But the problem you are running into is that it was a multifold set of
problems - which are often hard to untangle.  While I'm not sure how well
it worked in practice, the "justification" for the ARPAnet was to share
expensive resources owned by the USG and supplied to DOD/DOE contractors.
DOD and DOE were paying for lots of computing power at lots of places.  DOE
used almost anything they could get their hands on - particularly in the
scientific processing area, but the CS Research types had started migrating
to the PDP-10 for their specific serious work.   However, with the PDP-10,
there were N different OSs in use.  DARPA knows it costs the >>USG<< less
if the users operated with a DEC-supplied SW stack, but their CS researcher
seems to do more projects with more enhanced OS.  BBN has managed to get
DEC to pick up its own PDP-10 system and migrate the 'default' OS to be
based on theirs (FWIW: DEC is less impressed with ITS and WAITS in those
days for commercial reasons - I'm not going to go down that rathole). The
reader might try to remember that, as a general rule, DARPA and the rest of
the US government teams are trying not to fund what we might call "*core OS
Research."*

In 1983, DEC "discontinued" development and no longer offered for sale the
PDP-10 in favor of its now widely popular VAX series. DARPA switched to Vax
as the platform it will supply to its contractors (DOD and DOE, as well as
other depts, offered different systems).  However, with the vax as a common
platform for the DARPA contractor, there was still a need for some system
extensions like ports/sockets for different research projects DARPA is
funding.   The research community has started to switch to UNIX. But DARPA
is concerned about AT&T's "abandoning the OS on the doorstep" scheme.  So
the question was, how to get UNIX supported on Vax,  Since the version of
UNIX being used on the VAX by >>much<< but not all of the DOD and DOE
community was BSD, DARPA's solution was let a contract to create a support
group - CSRG was born.



> This story implies DARPA helped arrange Unix licences with the many
> defence contractors, albeit they only need binaries for BSD.
>
I did not imply that, nor do I think DARPA did. I think other parts of US
GOV did >>sometimes<< have access to the UNIX IP by means other than the
traditional license scheme from AT&T/WE Patent and Licensing group  -- i.e.
Otis Wilson *et al *(we have evidence of the same).  For instance, Ford Aero
was doing a joint project with AT&T for NASA [NASA is now an independent
agency, but I wonder if that was always true]. Ford Aero is known to have
had special access [and there seems to be evidence this was based on PWB
1.0 - which never was formally released outside of the Bell System].  There
have been other discussions that when other parts of Ford Motor wanted to
use UNIX, the Ford Aero folks were unable to help them.

We also know that Rand was an original (1960s) DARPA contractor [back to
its origin story as a research office inside the Air Force].   When the
folks from Harvard went to Rand and wanted to use UNIX, the first
commercial license was created by AT&T.  And we know that story.

There is evidence that some US government contractors, such as BBN, were in
a grey zone. I'll try to get some enlightenment from some of the BBN UNIX
folks I know. From discussions, it >>seems<< like the first version of UNIX
made its way into BBN and was part of a US Gov contract, probably shared
with ATT. But by the time of CSRG, BBN definitely had traditionally
commercial source licenses.   We also know that Ford became a traditional
licensee but started in a place different from others [particularly if the
reports of using PWB 1.0 were true -- that distribution was not available
from the AT&T/WE patent and license group].




> If the Internet Society’s ‘brief history’ is to be believed, Defence
> declared Unix a ’standard’ (for which work?) in 1980.
>
Please be careful here.  The IP based on the UNIX ideas was not a USG
standard for any department until FIPS-151 was published post IEEE P1003.1
- which was all in the 1980s. That said, there was often an *operational
standard *in many US government departments, including DOD, by the early
1980s; based on a preference for a flavor of UNIX by many users,
particularly researchers.

Furthermore, IP/TCP was the DOD's operational standard by the early 1980s,
but by the mid-1980s, DOD's DDN and DOC had picked ISO/OSI and GMAP
specifications over the IP family [and we all know how that played out in
the end].  A number of us in the industry at the time we scrambling how to
bring an ISO/OSI stack out on our products for the USG and the
Auto/Aerospace customers who were telling us they would not order equipment
without [and, of course, the folks in the EU were pushing X.25 and the rest
of ISO to counter IP's take off].    Metcalf's law would cause IP to win
out (i.e., economic reasons), but understand there is a difference between
an official standard and what was actually occurring.

[-- Attachment #2: Type: text/html, Size: 17502 bytes --]