Work in progress

In February 2014 I participated in the Coursera course “Internet History, Technology, and Security”. Part of it is an assignment: “Write an essay about how your first encountered the Internet or an earlier networking technology.” [sic]

I suspect that my history is a little atypical, so I'll keep a more detailed version here (the assignment is a maximum of 1000 words). The documents don't overlap completely, so potentially the version I submitted is of interest.

EXPAND: first encounter

In March 1979 my company sent two of us, a Mr. Hölscher (we didn't use Christian names) and myself to a data communications course at Tandem Computers in Frankfurt/Main. In those days, “data communications” definitely meant “point-to-point”. But while we were there, the Tandem support people, under Harald Sammer, were setting up the first European node of Tandem's internal network, based on their EXPAND network.

EXPAND is an interesting network because it literally expands the message-switching architecture of the Tandem system to a global scale. My understanding is that is was originally written for our customer McDonald's, a fast food chain. At the time McDonald's was barely known in Germany, but Tandem found it useful for its own internal systems as well.

An interesting implementation detail: the Tandem instruction set included an instruction to start I/O operations, called EIO (“Execute I/O”). For reasons I've never understood, EXPAND needed an indirect version of the instruction, “Execute Indirect EIO”. Considering the customer, it's understandable that its mnemonic was EIEIO.

The installation of the network was relatively uninteresting to us: the data communications course went in very different directions, programming applications to talk to each other over a point-to-point link. I had written a kernel debugger for Tandem's GUARDIAN operating system, and I needed permission to install it. The Tandem people agreed, and literally challenged me to compromise system security—a trivial matter, of course, with a tool that can modify kernel memory. The real question was how to do it undetected.

In the end, I changed the owner of my command interpreter (think “shell”) process to the super user, and then modified the password files. The super user was called SUPER.SUPER (the first part is the group, which is always unique for any user, and the second part was the user name) The group and user numbers were both 255, so status information would show something like SUPER.SUPER (255,255). I changed the user name for 255,255 to SUPER.TURKEY, and created a new user SUPER.SUPER (225,255) with the old SUPER.SUPER password.

It worked! The difference in numbers wasn't immediately obvious, and the network helped: they had added much additional security code to distinguish between local and network access, and they suspected that. Finally I had to show them.

It's interesting to think of what else was going on in the world at the time. In the USA, the most significant nuclear accident in US history took place at almost exactly the same time. In terms of networking, ARPAnet had been under development for nearly 10 years, and had been “operational” for nearly 4 years. Usenet was just an idea, and networks like BITNET (1981) and NSFNET were not yet on the horizon.

But there was one other network: X.25, developed by the CCITT. In many ways, it offered the same services as the Internet. But that's a long topic, so I'll talk about it some other time.

EXPAND: In earnest

The people at Tandem weren't overly upset by my antics in 1979. Two years later, the same Harald Sammer offered me a job, and on 17 May 1982 I started working with Tandem, just round the corner from where the previous incident happened.

Now networking wasn't just a curiosity: it was part of our working life. Email, in particular, transformed the way we worked. I was part of the Customer Service Organization, and we had support groups in Frankfurt, High Wycombe (UK), Reston (Virginia, USA) and Cupertino (California, USA). Our problem tracking system was a distributed database accessible from anywhere on the network. Even the personnel department worked on the network:

The email addresses here are composed of node name (\CTS in Cupertino, \SOSII in Frankfurt), the group name and the user name. It was a feature of EXPAND that there was a unified view of the entire network. This also extended to files, so there were messages like this one (which I have only as an attachment):

And network speed? The main backbones only ran at 56 kb/s. It was practical, but sometimes slow, to run interactive shells on systems in other continents, but we'd expect email to be delivered in less than an hour. Sometimes it was much faster, as this exchange (with Sue Kleiman at \SSG in Cupertino) shows.

The times are local time, and the message numbers are local. Normally Cupertino is 9 hours behind Frankfurt, but by coincidence this exchange happened when daylight savings time was still in operation in California, but had already finished in Frankfurt, so there was only 8 hours' difference.

What didn't we have? Graphics of any kind. And of course there was nothing like the World-Wide Web. But if you don't know about it, you don't miss it. We loved our network.

It's interesting to think of what else was going on in the world at the time. ARPAnet had been operational for nearly 7 years, and I don't know of any non-US node on the ARPAnet. BITNET had only just started, The Internet was just about to start operations. 4.1cBSD, the first operating system that enabled the Internet, was released in late 1982, and production networks didn't come until even later. In many ways that started with NSFNET several years later. We didn't feel like pioneers, but looking back, we were to some extent.

There's much more to say, but in short, on 17 May 1982 the network world became an important part of my life. When I found the Internet in late 1989, things changed a little, but not much.