[ih] Historical fiction

[Bernie Cosell]

On 10 May 2012 at 8:15, Sytel wrote:

> How was Arpanet viewed/seen *at the time*? How did it fit in the context
> of the era? What was its future envisioned to be?

I expect the various folk involved with it saw it pretty differently.  
>From my perspective, I didn't worry much about 'future' -- it was very 
hard to see quite what it would be good for but was an awful lot of fun 
to build..:o)

> What kinds of tests would be run on the Arpanet on a daily basis, 
> i.e.what would be a typical day in the lab? And who would be in charge
> of performing what kind of test? What would happen in the lab other than
> tests?

"lab"?  The only real tests I remember were ones orchestrated by 
Kleinrock at the NMC and I can't remember when they were run, only that 
we knew about them [of course] and were on more or less high alert to 
look for things breaking and such while they were testing.  Other than 
that kind of thing, I don't remember much 'testing', per se.

There was a lot of ongoing implementation and things changed over time 
[e.g., routing algorithm was redone at least twice in the early days].  
But the most remarkable thing to me, in retrospect, is the degree to 
which the damn thing *just*worked*.  As I look back over the listings 
[which I can barely apprehend any more] I realize that it was really 
quite a complicated system and it was pretty neat that basically it 
mostly just hummed along.

Things did break and fail a bit and so we had plenty of analysis, 
debugging and patching to do [Why'd imp 7 crash?  And we'd get someone at 
the site to help us poke around [via the switches] and we'd often plant 
some diagnostic code [to report things back to us], etc]  But overall it 
really just kept running and quite quickly the hosts connecting to it 
were *expecting* it to be there.

> What were some of the first things it was actually used for other than
> testing the capabilities of the network? When did it begin to perform 
> those duties?

Mostly in the early days, I think most of the 'testing' was via the 
hosts.  They'd hook up and try to get their host software to work or try 
some new application and we'd try to watch [from inside the IMP] to see 
what was going on and try to help figure out whether/if/what was going 
wrong.  While we were at it, we often discovered bugs and problems in the 
IMP code.  I can recall a few times when something that happened in the 
net hinted at a problem and we'd try to figure it out and fix it before 
anyone actually noticed it [on the theory that a bug not run-across by 
the users didn't happen... :o)]

> How did people react to its potential social aspects? Were they 
> evenconsidered at the time, or were they overshadowed by its potential
> as ascientific/knowledge-sharing system?

For me, I didn't worry much about that stuff --- we were involved with 
the hosts and the various applications and uses they were developing to 
try to make some sense, and some use, of the network [one I recall was 
using "raw packets" [is that right -- my memory is hazy] and an attempt 
[by Lincoln labs, I think] to use them to make a scrambled/digitized 
phone work over the net.  Those kinds of things often showed up problems 
in the code and so we would end up working with the host software 
developers and we all tried to make things work better.  But from my 
"bottom of the well" view, it was mostly an interesting, hard technical 
exercise -- new problems and challenges every day and with a finicky and 
complicted software system that had to be kept running *all* the time.

> What did one of the old packets actually look like? We haven't been able
> to do much except extrapolate from modern packets.

Boy, I don't remember -- the host packets you can get from BBN rpt 1822 
[and that info must be in the RFCs somewhere].  I've loaned out my copy 
of the IMP source listing but I'd guess that that's about the only place 
where you'd be able to find that info [I don't *think* we ever produced 
any external documentation of how the IMP-to-IMP packets were laid out]

> What were some common problems that happened during the early days of 
> testing? Other than the famous Login error.

I don't even know what the "famous Login error" was.  the most common 
problem was that a new host system would connect and then not be able to 
talk...:o)   I know that once when I was sending out a patch I 
fatfingered something [the details blur a bit with the mists of time :o)] 
and I took down the *whole* net: we were in the monitoring center and 
could see the net winking out, node by node.  I got the software fixed, 
but then we had to call EVERY site [only a dozen maybe] and get someone 
to reboot the 516 [at which point it would try to reload its software 
from a neighbor and would get the patched code and came back to life -- 
and so one-by-one we watched it come back.  As I recall it was early in 
the morning at BBN and so we had to talk a janitor or something through 
loading the IMP boot-tape into the paper tape reader at, like, 4AM 
pacific. 

> Was the system ever shut down at night? At which point did they stop 
> doing that, and what factored into that decision? At what point did the
> system become too important and decentralized to shut down *ever*?

AFAICM [new acronym: "... I can remember" :o)], the network was *NEVER* 
shut down.  Not once!  Even when folks were just trying new applications 
and developing software, almost from the first day the presence and 
functioning of the underlying network was taken for granted.  There were 
sites from Norway to Hawaii and so there was hardly an [eastern timezone] 
time that the network wasn't being used.

> What was the initial reaction to the first non-academic uses of the 
> network? Any steps that were seen as wrong turns? Any thatwere seen 
> aswrong turns at the time but worked out for the best eventually?

I thought it was fine.  Note "first non-academic" is a bit odd -- it 
implies a more direct path of development than I think actually happened. 
To my memory, there were a lot of experiments [e.g., Bob Thomas's 
distributed OS stuff and the scrambled-phone I mentioned earlier] but it 
was all not [IMO] pointing anywhere -- it didn't feel like it'd be useful 
to third parties: it was all interesting software [often difficult] and 
everyone was learning a lot about what the net could and couldn't do and 
such, but just as for the PC what really was needed was a "killer app".  
And of course, that was email.  email, IMO, brought the first 
non-technical people onto the ARPAnet and changed things [instead of 
system programmers calling when there were problems, there were office 
staff people and such].  And it was all downhill from there...:o)

[in fact the 'academic' is a separate story best elaborated by the CSNET 
folk [I was only perphierally involved in that] -- the ARPAnet wasn't an 
_academic_ network: it was a DoD research network and as such only sites, 
mostly universities, granted, were allowed to connect.  And were 
*supposed* to limit their use to whatever project they were working on.  
Among other things, email changed all that, of course..:o)

  /Bernie\
-- 
Bernie Cosell                     Fantasy Farm Fibers
mailto:bernie at fantasyfarm.com     Pearisburg, VA
    -->  Too many people, too few sheep  <--