[ih] Saving IETF history

[Jack Haverty]

The listing I analyzed was from 1973.  By then it was no longer
necessary for any human intervention.   If the IMP crashed, or simply
became nonresponsive, the hardware would force it to jump into a set of
instructions that sought out a neighbor and asked for a reload.

As part of that patent effort, I analyzed the reload-related parts of
the IMP code, instruction by instruction, and the interactions between
the 2 IMPs involved.   I suspect it violated more core principles of
layering, software engineering, coding structure, and other such rules
than any software had to date.   The part where the network I/O
overwrote the memory while the processor was executing from it was
especially clever.

Most importantly, it worked.

/Jack


On 5/12/21 6:06 PM, Dave Crocker via Internet-history wrote:
>
>> "downloading new versions of system software".
>
> Arpanet coming out party, Fall, 1972, bowels of the Washington DC
> Hilton.  I was asked to demo the net to Lipinski, Sr., of the IFF.  I
> connect to BBN and run something.  Then connect to ISI and run
> something.  He's thoroughly bored.  Too easy, quick and unengaging.  I
> could tell that he was not appreciating what this was..
>
> Then the TIP crashes.  We wait a moment and he asks how long before we
> can continue.  I check and am told a reload is needed.  I tell him
> that and he says he'll go off to a conference session until the system
> is reloaded.  I explain the delay will only be a few minutes.
>
> He says that can't be.  They haven't even rolled out the paper tape
> reader yet.  I explain they won't be doing that, and tell him that
> Boston will be directing a download over the net from a neighbor.
>
> This is a bright, experienced guy.  He takes in what I just said and
> THEN he finally understands the net.
>
> d/
>