[ih] Arpanet physical connectors

[Louis Mamakos]

On 23 Jul 2020, at 11:53, Noel Chiappa via Internet-history wrote:
> The VDH interface used a modem, and I have this impression/memory that 
> the
> IMP interface to that would have been the stock IMP/modem interface. 
> What was
> on the host end would have been whatever was stock for the modem 
> interface
> used. On the IMP end, I have a vague memory (from a day when we shut
> everything down to reorganize the rat's nest of cables around the 
> IMP+TIP)
> that it was a hairy connector in which each pin was a coax job, with a 
> ground
> shell around a center confuction - but don't depend on that memory. 
> The Bell
> 30x modem manual may give the details.

Well after the ARPANET/MILNET split, there were a number of IMPs 
installed
as part of an NSFNET interconnection project.  We had one installed at 
the
University of Maryland at the time as part of this, interconnecting the
ARPANET to the Phase-1 (Fuzzball-based) NSFNET and Phase-II NSFNET 
routers
that were also colocated other parts of the same computer room.

At that time, BBN had insisted we use the X.25 interface on the IMP, 
presented
in the form of a 56K synchronous serial interface on V.35 connector.  
This was
cabled to a synchronous null modem to provide clocks, with the other end 
intended
for a MicroVAX-II.

We had received a very early version of an X.25 interface for the Q-bus 
on the
MicroVAX-II, with a V.35 interface.  It was an early (prototype?) board 
and the
designers of the board managed to get the pads for the V.35 connector 
mirror
imaged on the PCB.  So we had about a 1 foot long adapter cable with a 
V.35
male on one end and V.35 female on the other, with the wires twisted 
around to
under the layout problem.

Of course we tried to crank up the clock on the synchronous null modem 
to see
how fast it would go.  I don't recall if it was the IMP/PSN or the Q-bus
interface that tapped out first.

Turns out that X.25 was a bad choice for this application.  The X.25 
interface
on the MicroVAX would need to signal a switched VC to the destination 
IMP/port
to send packets to it.  As you might expect, a router between the 
APRANET
and NSFNET had.. quite the spatter of destinations on the ARPANET 
getting traffic.
To the point that you'd start to thrash the limited number of 
simultaneous
connections that the X.25 board would support.

There was a later version of that board from ACC that worked out a bit 
better.
I do recall that the BSD UNIX kernel driver for that board really was 
quite
the thing to behold in all its complexity.

On another note, the University never did get back its "loaner" 
LSI-11/73
CPU board used to upgrade of of the MILNET/ARPANET "mail bridges" from 
the
LSI-11/23 CPU.  Something about procurement processes and the Internet
about to explode, again..  Fortunately, it had no asset tag on it and it
wasn't missed.

louie