[ih] ruggedized Honeywell 516 ARPANET IMP cabinet top lifting hooks (Was: The IMP Lights story (Was: Nit-picking an origin story))

[Leonard Kleinrock]

Two side points:
1. When Larry gathered us to write the Arpanet spec for the RFP, we decided that the reliability specification would simply be that the topology would be 2-connected (that is the the topology would provide two independent paths between every pair of Imps) thus ensuring that if any single Imp or link went down, all the other surviving Imps could still communicate.
2. I recall seeing the new Honeywell DDP 516 demonstrated at what I believe was the Full joint computer conference of 1968. It was suspended from the ceiling with wires connected to the hooks on top of the Imp. 
Len
Sent from my iPhone

> On Aug 25, 2025, at 10:01 AM, Jack Haverty via Internet-history <internet-history at elists.isoc.org> wrote:
> I heard similar claims about the ARPANET's survivability.
> 
> IIRC, the IMPs themselves were militarized, but could still be obliterated in a "hostile battlefield".  But the ARPANET itself would continue to function, with the remaining undamaged IMPs still operating and providing host-host communications - even if enough nodes had been destroyed to render the network into several fragments.   The reasoning was that, unlike other network designs of the era, there was no central locus of control in the ARPANET.
> 
> In other network designs, there was a control center located at some node, which handled setting routes, establishing and closing connections, et al.  So if that node was destroyed, all nodes would become unable to communicate, even if they were undamaged.   IIRC, IBM networks of that era had that characteristic.
> 
> IMPs had a NOC, but it wasn't involved in minute-by-minute control operations.   Each IMP was self-contained and able to interact with whatever other IMPs and circuits it could contact after events like circuit failures or disappearance of other IMPs.  IMPs didn't need a NOC to open/close connections and pass traffic between hosts.
> 
> IMPs themselves wouldn't survive a "nuclear event", or likely even a non-nuclear one.   But the remaining IMPs would continue to operate the ARPANET even if it was split up into several pieces.   IMPs wouldn't survive, but the network would.
> 
> The Internet followed a similar design principle, with no centralized control - although we thought about it when TCPV2 was being transformed into TCP/IPV4, as a possible approach to providing functionality such as "policy-based routing".
> 
> Jack
> 
> PS - The "Pluribus IMP" was a later design, using a highly redundant multiprocessor computer architecture that could survive multiple failures, although not a bomb.  Probably.  It was used in some "clones" of ARPANET within the government.  I once heard a rumor about when one such IMP was decommissioned.  The technicians decided to see how robust the hardware really was.  So they "decommissioned" a node, while it was running, using an M16 rifle.  It took way more shots than they expected before the IMP finally stopped working. Apocryphal story?  Maybe.  I wasn't there to watch.   More info about Pluribus IMP at https://apps.dtic.mil/sti/html/tr/ADA151312/index.html
> 
> PPS - re the massive hooks on the original IMPs:  IMPs were built using commercially available computers from Honeywell.  I always assumed that the militarized model had hooks (and other features such as those massive connectors) because it was a requirement for DoD procurements, and Honeywell computers were probably used in the military for other purposes than IMPs.  Hooks were probably required for use in moving equipment on/off naval vessels with cranes, securing equipment when used in places while in motion (ships, jeeps, planes...) and other such needs in military installations.
> 
> 
> On 8/25/25 08:24, Dave Crocker via Internet-history wrote:
>> On 8/25/2025 6:00 AM, Steve Crocker via Internet-history wrote:
>>> There was no requirement or expectation that the IMPs would survive a
>>> nuclear event.
>> 
>> 
>> Again, as a mere passenger on the packet-switching technical development train, trying to learn what I could, I don't recall ever hearing nuclear survival as a goal, but I do remember repeatedly hearing something like "survive hostile battlefield conditions".
>> 
>> When reciting this to others, such as doing TCP/IP presentations, I noted that this turned out to include average commercial operating environments...
>> 
>> Also, my understanding is that a precise example of hostile battlefield conditions was not tested until the first Iraq War, and then it was Iraq's network (using Ciscos?) that did indeed survive...
>> 
>> d/
> 
> -- 
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history
> -
> Unsubscribe: https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history