[ih] Failures of the early Internet

Tue Jan 23 08:24:19 PST 2024

The Div 6 VAX Unix project evolved over time and people and ran into the
late 1980s.  It was typically a one-person project and, at the end, I was
its supervisor.  As I recall, the progression of people working on it was
Dennis Rockwell, Bob Walsh, Karen Lam and then David Waitzman.  Again,
relying on memory, Bob and Dennis worked on evolving Gurwitz's BBN 4BSD
code, porting it to 4.1 and 4.2 BSD. Recall the Berkeley 4.1c/4.2 TCP was a
rewrite of the BBN 4.1 TCP code by Bill Joy -- exactly to what degree Bill
Joy did his own thing and how much he relied on the BBN code is
occasionally debated, but I do know as late as 1988, Berkeley would defend
a bug in the BSD TCP code by pointing out it originated in the BBN code, so
obviously Bill referred to the BBN code.  In any case, DARPA continued to
support the BBN version for a few years after 4.2BSD came out -- and, if
memory serves, BBN licensed its version to some of the rising desktop
server vendors (e.g. Apollo).

I believe Karen worked primarily on enhancing the BSD 4.2/4.3 TCP, but I
could be wrong.  I know David worked closely with Steve Deering to enhance
4.3 BSD to support multicast.

I am probably getting some details wrong.

Craig

On Mon, Jan 22, 2024 at 6:57 PM Jack Haverty via Internet-history <
internet-history at elists.isoc.org> wrote:

> Yes, there was TCP work in both pieces of BBN.  At first, Bill Plummer
> was doing Tenex on the PDP-10, and I was doing the first Unix
> implementation on a PDP-11/40.  Both of those were involved in the "TCP
> Bakeoff" that Jon Postel organized.  Later more implementations were
> done, e.g., Charlie Lynn in Div4, and in Div6 Gurwitz (Vax Unix), Hinden
> (TAC), Sax/Edmond (HP3000 Unix), Wingfield/Nemeth (11/70 Unix) and maybe
> some more I've forgotten now.   For the historians, lots of detail in
> the BBN QTRs of the era, accessible on DTIC.
> Jack
>
> On 1/22/24 17:09, Barbara Denny via Internet-history wrote:
> >
> >
> > I am pretty sure Charlie Lynn was working on a TCP implementation in Div
> 4 (Packet Radio group) when I was  there.  I did figure out eventually that
> the gateway work was in Div 6.
> > barbara
> >      On Monday, January 22, 2024 at 03:28:15 PM PST, Jack Haverty via
> Internet-history<internet-history at elists.isoc.org>  wrote:
> >
> >   Ahah!  Another piece of history falls into place.
> >
> > I recall Vint asking me at some point if I'd be willing to take over the
> > gateway work and turn the core Internet into a 24x7 operation, as the
> > Arpanet had become by then.  I didn't have a clue how to do that but it
> > sounded like a fun project so I said "Sure".   That may have occurred at
> > the same Internet meeting that Noel recalls.
> >
> > The gateway work moved shortly thereafter, from BBN Div4 (where Packet
> > Radio work happened) to BBN Div6 (where Arpanet and various TCP
> > implementations were done).
> >
> > I didn't find out until years later that some of the Div4 folks were
> > sure that I had "stolen" their project.   That wasn't true, but I was
> > blamed for it.
> >
> > Now we know the Truth.   Noel was the culprit!
> >
> > Jack Haverty
> >
> >
> > On 1/22/24 15:11, Vint Cerf via Internet-history wrote:
> >> While I don't remember this specific incident it sounds authentic (I
> often
> >> write down action items to remember them).
> >> It may be that I was intending to move into an operation posture, in
> which
> >> case I would want the same division that was monitoring the Arpanet
> IMPs to
> >> be monitoring the gateways.
> >>
> >> v
> >>
> >>
> >> On Mon, Jan 22, 2024 at 5:55 PM Noel Chiappa<jnc at mercury.lcs.mit.edu>
> >> wrote:
> >>
> >>>        > On Fri, Jan 19, 2024 at 9:20 PM Jack Haverty wrote:
> >>>
> >>>        > Anybody else have recollections of early failures...?
> >>>
> >>> This may not be the _kind_ of failure you're thinking of, but I
> remember
> >>> one...
> >>>
> >>> We had just brought up the first ARPANET-connected gateway at MIT, and
> we
> >>> needed to let the BBN 'core' gateways (I forget what the jargon name
> for
> >>> them
> >>> was at that point) know where MIT (net 18.) was connected, now.
> >>>
> >>> So, I got a copy of the GGP spec (IEN-30, or IEN-109), but I was too
> >>> busy/lazy to do a full implementation of GGP - I just typed in (in
> octal!)
> >>> a
> >>> 'routing update' packet that showed MIT-GW as connected to net 18,
> along
> >>> with
> >>> a few lines of code to send a copy, once a minute, to a list of
> gateways on
> >>> the ARPANET (here:
> >>>
> >>>      GWTAB:.BYTE  12,3,0,50              ; BBN
> >>>            .BYTE  12,2,0,31              ; CSS-GATEWAY
> >>>            .BYTE  12,1,0,24              ; DCEC
> >>>            .BYTE  12,3,0,110              ; RCC-GATEWAY
> >>>
> >>> is a copy of a later version of the table). I loaded this in to the
> MIT-GW,
> >>> started it up - but it didn't seem to be working.
> >>>
> >>> Shortly thereafter, IIRC, I got a call from someone at BBN, saying
> >>> something
> >>> like 'are you sending us GGP packets?' Apparently all their gateways
> had
> >>> crashed - after they got my routing update.
> >>>
> >>> It turns out they had changed the format of routing updates - but
> nobody
> >>> had
> >>> gotten around to updating the GGP documentation. So I had been sending
> them
> >>> old-format routing updates - which gave the BBN gateways indigestion.
> >>>
> >>> I remember that at the next Internet meeting, I wound up telling Vint
> this
> >>> story (probably because somebody had been giving me grief about
> 'crashing
> >>> all
> >>> the BBN gateways'), and he got this look on his face, and pulled out
> his
> >>> little 'things to do' notebook and wrote something in it.
> >>>
> >>> We were later speculating that he'd written down 'fire Div 6' (or Div
> 4 -
> >>> whichever one it was that Ginny worked for), because shortly
> thereafter, it
> >>> was announced that Div 6 (or Div 4, whichever one it used to be) would
> be
> >>> replaced by the other one in running the BBN gateways. I have no idea
> if
> >>> there's any truth to that, but the change was made shortly afterwards!
> >>>
> >>>            Noel
> >>>
> >>> PS: A later version of the offending code still exists; here:
> >>>
> >>>      http://ana-3.lcs.mit.edu/~jnc/tech/gw/conf/sbr/route-1.mac
> >>>
> >>> if anyone is interested. Note that the update format there is _not_
> the one
> >>> in IEN-109! :-)
> >>>
> >>
>
> --
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history
>

-- 
*****
Craig Partridge's email account for professional society activities and
mailing lists.