[ih] Fwd: Arpanet Reconstruction Project - an update

[Vint Cerf]

sharing with permission
v


---------- Forwarded message ---------
From: Oscar Vermeulen <oscar.vermeulen at hotmail.com>
Date: Tue, Feb 3, 2026 at 2:35 PM
Subject: Arpanet Reconstruction Project - an update
To: Robert Kahn <rkahn at cnri.reston.va.us>, Steve Crocker <steve at shinkuro.com
>
Cc: o.oosterwijk at ceds.dev <o.oosterwijk at ceds.dev>, Leonard Kleinrock <
lk at cs.ucla.edu>, Vinton G. Cerf <vint at google.com>, Lars Brinkhoff <
lars.brinkhoff at gmail.com>, Charley Kline <csk at mail.com>


Dear Sirs,

We bothered you with a few emails last summer, about our "Arpanet
Reconstruction Project" - and we promised to keep you informed.

I hope you do not mind us contacting you again for such an update. But we
have made quite some progress, enough to send you a link to a working
Arpanet recreation:

https://obsolescence.dev/arpanet_home


This replicates a fully functional Arpanet with the topology from May 1973,
with 35 IMP emulators connected, running their 1973 firmware. Accessible
currently are the three MIT PDP-10s running ITS, and the Stanford SAIL
PDP-10 running WAITS. The web page contains 8 web terminals, labelled as
'seats at the 1972 ICCC conference' where visitors can connect through the
Arpanet IMPs to one of these four systems and attempt a few of the "Arpanet
Scenarios" from the event's booklet.


   - And this begs the question, as computer history hobbyists, did our
   tone of voice (aimed at a general audience) capture the reality of the
   event? We decided to shape this web page (the prototype for a museum
   exhibit) around the idea that the 1972 ICCC conference was a breakthrough
   event, and it is interesting to experience 'what it was that made people
   see'. But we base ourselves on reading up decades after the event. It will
   be very welcome to hear if perhaps we overly romanticised the event - and
   should amend the "storyline".


The web site with arpanet terminals acts as the prototype for:

   1. An interactive museum exhibit - we dream of having some nodes located
   at their historical locations
   2. An open source 'arpanet in a box' package that people can run on
   their own machines and explore/enjoy
   3. A hobby Arpanet that our group can connect to with their replica
   PDP-10s and -11s.


This is only work in progress - the text of the page, the exact commands to
relive the Scenarios - all of that will need a small group of volunteers to
perfect before we would consider this a public site. But the setup is
properly working, and the ITS and WAITS systems are rich enough in context
that indeed, you will see students' messages on the Stanford machine from
the early 70s; and ITS user directories with some well-known MIT users and
their software projects. In other words, we believe this can truly be a
realistic experience of the Arpanet circa 1972-3.

Work is underway to reconstruct the TENEX system, and files from BBN and
the University of Utah have been found to make these nodes more than
sterile 'bare OS' systems. But the OS code is not trivial to get to a
working stage, it will take some time. Also, work is underway to present
the UCLA SDS Sigma 7 online with an operating system that is maybe not the
period-correct one, but at least a modified CP-V that will present itself
in a similar fashion. Also, we have the BBN Network Control Center on the
to-do list. Of course, the IMP network already sends data to BBN's IMP node
5, and we are processing the data that this particular IMP sends to its
host 0. But the end goal would be to have a replica PDP-1 handle the data
flow. We also have high hopes to get the two original Multics machines
online, a group of people has worked on them and the challenge will be to
complete the missing NCP code for Multics. But that is now feasible - it
will be done. Lastly, a period-correct PDP-11 Unix v6 with NCP connection
sounds like it will be on the cards later this year.

We'll also have a hardware replica of the H516 IMP ready as open-source
hardware in the coming months. It is needed for museum exhibits, and
hobbyists will enjoy it as well. We should not over-represent its
importance: behind its replica front panel will hide a Raspberry Pi with a
Honeywell 316 simulator running. But it will be fully functional, connect
over serial lines or internet UDP connections to 'the reconstructed
Arpanet' and be a nice functional museum exhibit, as we hope.

(You might have forgotten, our series of replicas is at
https://obsolescence.dev).

I hope you do not mind our longish update above - too much information, I'm
sure. But feedback of both the positive and negative kind is very welcome.
It will help us get the historical context right. Please keep in mind
though, that the web pages' text are only drafts. We focused on getting the
actual Arpanet simulation working reliably first.

Kind regards,

Oscar Vermeulen
Cc Lars Brinkhoff
Cc Otto Oosterwijk



-- 
Please send any postal/overnight deliveries to:
Vint Cerf
Google, LLC
1900 Reston Metro Plaza, 16th Floor
Reston, VA 20190
+1 (571) 213 1346


until further notice