[ih] Significant milestones in the history of TCP/IP

[John Day]

That would sort of depend on how much your site was relying on the Net for its work.  

On any given day in the early-mid 70s, we would have anywhere from 20 - 30 people on the Net working (the terminals would have been in the building and running at 9600). Of course, if you were working from home which several of us did mainly evenings. Dial-up was slower. Most were developing code (editing remotely, no local file system) or related tasks.  Several times a day we would download code files to test. Others were submitting big compute jobs at either CCN or Rutherford and either doing sizable print jobs or generating large files for Calcomp plotters. In addition, we had users of a land-use management systems that used multiple databases on the net and displayed it as shaded maps on the LSI-11 based terminals that had plasma screens and touch.

I remember early on when we were transitioning we had a slow line (perhaps it was 4800, might have been slower) just for submitting jobs and downloading code files. It was very tedious.

Without the 56K lines, I think that would have been very painful.  As it was, the pain was more how slow the hosts were rather than how slow the Net was.  Which was probably what was important: that the hosts were more a bottleneck than the network.  Not that one would saturate the lines. That seldom happened.  But slow lines with character echoes would have been very painful.  It wasn’t great as it was.

Take care,
John

> On Sep 17, 2015, at 13:11, Noel Chiappa <jnc at mercury.lcs.mit.edu> wrote:
> 
>> From: John Day
> 
>> The line speed was very important.  Remember the ARPANET was built to be
>> a production network ... The Net would have worked, but in terms of
>> Robert's goal of a resource sharing network that would allow sharing of
>> major computing resources, it would have been painfully slow, and
>> probably deemed a failure. The 56K lines kept the perceived response
>> within expectations. ... Even those days, 9.6K was reasonable speed for
>> one person (maybe two) at a terminal doing development.  Sharing that
>> with a few hundred with more than a few doing essentially file transfers
>> for RJE and you have one painfully slow network.
> 
> I am less certain of this than you are. I had the experience of working for
> quite a while with Proteon, where the only link to the rest of the Internet
> was, for some time, a 4800 (or maybe it was 9600, the memory dims, alas) baud
> line. I don't remember it being really painful.
> 
> (Admittedly, the line did have some pretty fancy header-compression on it,
> using an algorithm from Dave Reed. If there were more than one connection
> using the line, one didn't get the maximal compression, but even then you'd
> get some compression; e.g. the source address - for outgoing packets - was
> mostly, or entirely [remember, this was the era of timesharing machines, we
> only had one :-] the same, other fields were the same, etc.)
> 
> Your point about lots and lots of users (the Proteon user community was
> admittedly fairly small, initially, at least - a half dozen or so) sharing a
> line is taken, but at the same time, there would have been a lot more lines,
> spreading the load.
> 
> I suspect the only way to say with any certainty how well a network built out
> of lots of slow lines, as opposed to a few fast ones, would have worked is a
> comprehensive simulation. Which is not likely to happen, of course! ;-)
> 
> 	Noel