This is my second blog entry, still recounting my history.
History – part 2
My story continues in 1980 when I was a tenured lecturer in Computer Science. At that time I had become interested in networks and the emerging internet in particular. Australia had a fragmented academic network at the time with the public X25 network and CSIROnet the only long haul options, with some use of dialup modems. It was also the time when Unix was starting to takeoff in Universities - having arrived in 1975 but now becoming suitable for large scale teaching.
At the end of 1979 the Basser Dept of Computer Science at Sydney installed a DEC VAX780. Through our close relationship with the Bell Labs Unix group we were able to use the latest Version 8 Unix and our programming staff (led by Piers Lauder) did extensive work to make it handle large numbers of dumb terminals in a student teaching environment. Looking back on it this was amazing work. The machine was powerful for its day but less powerful than the Apple watch of today! It had 4Mbytes main memory (no typo - that was MEGA bytes) and a 1MIP processor. Yet it handled up to 100 student terminals running a text editor, compiler and utilities. At 100 it was only just usable but at 60 terminals it was fine. Part of the secret was the development of an inovative scheduler (the “fair-share” scheduler) that scheduled processes on the basis of resource usage (cpu/mem/IO etc) rather than just CPU, and had an approach that shared the resources equitably over of long period (eg days). This work (mostly by Piers Lauder and Judy Kay) is the foundation for fair-share scheduling right up to the present day.
The early 1980’s was also the beginning of cheap academic networking in Australia. Fixed line modems were cheap and accessible allowing a simple async character connection between two machines. Our first such connection was a 2400 bps line to UNSW where a lot of Unix development was going on. This allowed remote interactive sessions and our programming staff could work with UNSW programmers on Unix development. The missing service was the ability to send email and files. Unix had a primitive system called UUCP (Unix to Unix copy) but it was direct from one machine to another without any routing capability. Routing of messages was performed manually by embedding the route in the address. This was primitive and inconvenient and inspired me to propose a system that exchanged routing tables and did the routing automatically. With Piers Lauder the first version was developed (called the Sydney Unix Network or SUN) and deployed around Australia to create the Australian Computer Science Network or ACSnet.
Things really took off at that point with all University CS departments joining the network and new versions of the system produced with more facilities. Eventually we commercialised the system (renamed as MHSnet - more on that later) and continued to develop it for many years after.
The key architecture of MHSnet is that it was a store-and-forward message system with a very flexible transport layer that could run over any sort of underlying communication link, and a flexible application layer that could provide message services for things such as email and file transfer among many. The transport layer could run over dialup, fixed lines or services such X25 and the emerging TCP/IP networks. The core of MHSnet was a heirarchical message routing service that exchanged routing tables between neighbouring nodes but automatically pruned the table where possible.
On a personal note, the 1980s were when my incredible children were born - Sarah in 1980, Rebecca in 1985 and Jonathan in 1988. All now with PhDs and in senior positions in research and academia.