University Networking
From ZaInternetHistory
| Table of contents |
Overview
This section endeavours to give some background information about the developments of university networking in the 1980's. This is necessarily incomplete, nothing was centralised, there was no forum for exchanging information and ideas, so no one had the overall picture. That in itself describes pretty accurately the mess that things were in.
This was still the era of the mainframe in South Africa. IBM PCs were around, Apple ][ computers were around, these were to be found mainly in student labs. Any networking that was to be done would have to be done via the mainframes.
There was a variety of computer manufacturers that sold computers to the universities in the 1980s. Names like IBM, ICL, Burroughs, Sperry/Univac, DEC/VAX, Hewlett Packard, Control Data, Data General, Prime and others all had some of the business. There were geographic regions of influence. In the Transvaal, IBM seemed to hold sway (hence the SNA network linking Potch/Wits/Pretoria/CSIR. In the Western Cape, there were VAXes linked between the Universities of Cape Town (UCT) and Stellenbosch (SUN). Hewlett Packard seemed to have the business in the Natal area, although there were VAXes as well. In the Eastern Cape, Rhodes had a Control Data Cyber, the University of Port Elizabeth had a Burroughs.
Even after the Uninet project had been running for some years, this difference in the mainframes (which implied differences in networking protocols) still dominated. Eg, at a Uninet conference in October 1988, Mike Lawrie presented the only paper on the TCP/IP protocols and the benefits thereof. A much later Uninet Board meeting took Mike Lawrie's advice that "Uninet should get to the standards of the Internet and follow those standards as they evolve". This acceptance put to an end the protocol wars that were smouldering. That differences of opinion should exist was hardly surprising,. In the early 1990's there were protocol wars in the USA about the benefits of TCP/IP vs OSI.
To the best of the Mike Lawrie's knowledge, the wide area university and research networks that existed in South Africa in the 1980s were these. Corrections and extensions to this list will be very welcome.
- The CSIR had a network that linked IBM and Cyber computers in Pretoria to research units across the country. This used Case multi-protocol multiplexors. It did not link the IBM to the Cyber, though.
- Potch/Wits/Pretoria/CSIR, and IBM SNA network using 9,600 bps leased lines.
- UCT/SUN, a Decnet network linking VAXes
- University of Natal, linking HP computers at the Durban and Pietermaritzburg campuses.
- ... others ...
CUP Computing Subcommittee
The Committee of University Principals (CUP) was a body that was formed (by statute) to oversee the interests of universities in South Africa. In the 1970's, growing out of efforts by the university IT staff for a forum to discuss common issues, the CUP formally established a computer subcommittee. This subcommittee, at a meeting at Rhodes University on 6th June 1986, formed a networking subcommittee. This was chaired by Philip Welman of Potchefstroom University, the other two members being Willie Fryer of the University of Pretoria and Mike Lawrie of Rhodes University. This committee met on several occasions, but with the best will in the world could not produce ideas for solving the "protocol war" problems, nor produce acceptable proposals for installing and running a network because of the high costs of Telkom (http://www.telkom.co.za/) circuits. With the best will in the world, the suggestions were to start with the RSCS protocol in a BITNET-style interconnection, and then migrate to the OSI standards. Not a word of TCP/IP.
Uninet Starts Up
In 1987, the President of the Foundation for Research Development, Dr Rein Arndt, took an active interest in the formation of a computer network. This was done in consultation with interested parties, including the CUP's computer committee (at a meeting at RAU, (when?? 1988??)), and with the networking subcommittee. In due course, seed money was made available by the FRD to purchase networking equipment and to hire circuits. One of the motivations for a network was that South African scientists had a need for access to a supercomputer, and without a network there was no way to make such a computer accessible to the scientists. As an aside, it is interesting to note that some ten years later there is still no supercomputer for these scientists.
Vic Shaw formerly of the CSIR, was appointed to the FRD as manager of Uninet in October 1987 (?) to co-ordinate the efforts of the computer staff at those universities that took an interest in establishing a national research network. A study into networking needs was conducted, and surprisingly while it mentioned the TCP/IP protocol, that protocol was not identified as even being the way forward. This may well have been a byproduct of the sanctions that had isolated South Africa from the mainstream of scientific developments elsewhere in the world.
One of the problems facing Uninet was that of Telkom (http://www.telkom.co.za/) regulations. The "third party traffic" rule was notorious, and had been misused time and again to close down such things as PC based BBSes. The rule applied only to leased lines, not dialup connections, and in a nutshell prohibited traffic from organisation A from being transferred across a leased circuit that linked organisations B and C. Clearly, Uninet needed to do this very thing when it linked up the various research and academic institutions of the country. Vic Shaw applied to Telkom (http://www.telkom.co.za/) for a concession under an escape clause about "common interest group". This application was made on 7th January 1989, and permission was granted in due course.
The networking equipment that was purchased for Uninet was Case multi-protocol multiplexors, which had been used very successfully on the CSIR's network. The circuits that linked key hub sites of Uninet operated at 64 Kbps on Telkom (http://www.telkom.co.za/)'s Diginet network. Uninet was granted special permission by Telkom (http://www.telkom.co.za/) to operate this network under the "common interest group" escape clause of Telkom (http://www.telkom.co.za/)'s "no third party traffic on leased lines" regulations.
All of this took time to get into place. (When were these circuits operating? They were in place by Feb 1989 when the Rhodes Fidonet gateway started working).
The muxes of Uninet provided virtual circuits on top of the physical ones. Any port on one mux could be allocated synchronous or asynchronous bandwidth to any port on any of the other muxes. Muxes were located at all of the sites that wished to participate in the Uninet project. These sites included UCT, Rhodes, Stellenbosch, Potch, Computer Centre for Water Research, University of Natal Durban, Wits and a few others (which??).
It was now possible for the VAXes at UCT and Stellenbosch to communicate with, say, the VAX in the Physics department at the University of Natal Durban by means of a Decnet channel, and the IBM at the CCWR to join the network of IBMs at Potch, Wits, Pretoria and the CSIR. No one had a ready solution for the problem of getting the VAXes to communicate directly with the IBMs, for example, quite why the JNET package was not run on the VAXes defies any understanding on the part of Mike Lawrie.
The only commonality within the early days of Uninet was that the Cyber at Rhodes had a mechanism to exchange email with just about any of the computer brands that existed on Uninet, and it soon became a key element in what was to follow.
Six times through Wits, or bust
There was a big difference between physical and virtual routes of the dataflows across the early Uninet network.
By 1989, Rhodes was acting as a protocol converter gateway for email of Uninet. The Cyber at Rhodes could handle the RSCS protocol, and thus could exchange email with the IBM hosts of Uninet, and at the same time it spooled email into a file to be collected by a program like kermit for hosts like the Hewlett Packards and VAXES in Natal and the Western Cape. All of these connections, including the kermit ones, ran across virtual circuits of Uninet.
It so happened that the route taken by email between the HP on one floor of the computer building at the University of Natal Pietermaritzburg and the IBM mainframe in the basement of that building was very convoluted. Looking at the virtual circuits, the HP communicated "directly" with the Cyber using kermit protocols, the Cyber accepted this email and converted any that was addressed to the IBM host into the RSCS protocols, and sent it "directly" to the IBM at Pietermaritzburg. Straightforward, until one looks at the physical circuits. From Pietermaritzburg, the 64 Kbps diginet circuit went to Wits. Potch was on a stub link to Wits. Rhodes was on a stub link to Wits. So, the kermit virtual circuit between the HP at Pietermaritzburg went physically via Wits. The Cyber then converted it to RSCS protocols and sent it to Potch (via the physical circuit to Wits). Potch sent it on its physical circuit back to Wits by means of the SNA protocol, and Wits sent it to Pietermaritzburg. Well, it was something like that, but a bit more complicated because among other things there was more than one mux at Wits.. All told, we counted six transits of the email through Wits in order for email to flow from one floor to another in the same building at Pietermaritzburg.
Search for an Internet Gateway
In spite of convoluted physical routes, Uninet worked well. Email exchange between hosts run by separate autonomous institutions was taking place. This was indeed a very big step, but the bigger need was for email to flow between Uninet and the Internet. The efforts to get an international link for Uninet started in earnest in 1989. The problem was political far more than technical, but that is not to say that the technical problems were trivial.
Some time before 1989, UCT had established a UUCP connection to UUNET in the USA. This ran across the international X.25 service of Telkom (http://www.telkom.co.za/). Unfortunately, after about three weeks the folk at UUNET pulled the plug on the grounds of South Africa's apartheid policy. They felt that they could not be seen to be co-operating in any way with a South African organization, no matter how liberal that organization might be. A number of individuals, including the author, had private and personal signons on hosts in the USA and Europe, granted out of a spirit of generosity. A researcher at the University of Natal Durban, Dave Levy, had connected the VAX there to the NASA SPAN network on this basis, and it was operating in 1989.
Letter from Bitnet
At about the time when JNET was being brought up to speed to link the Cybers in the Computing Centre to the VAX in Physics, the author wrote to the Bitnet organisation and requested a connection. A simple request, but one has to bear in mind that sanctions against the country were in full force. Even though it was not against the USA's Comprehensive Apartheid Act to deal with South African academic institutions, it would have to take a very brave institution in the USA to deal with any South African university. Organisations like SPSS Inc simply "pulled the plug" and refused to renew the license for a liberal anti-apartheid university like Rhodes to run their software.
That said, an Elizabeth Kilkoyne of Bitnet wrote back saying that we could connect to Bitnet. The author could not believe his luck, but could not afford the requred 9,600 bps leased line to the USA. The matter was taken to the networking group (?? Cannot recall whether this was the then Uninet Board or the CUP networking subcommittee, but Vic Shaw was present), who in effect laughed this out of court on the grounds that this was permission granted at a "technical level" and still needed the approval of the Bitnet governing body. There were too many people in that committee who believed that the governing body would reject the matter. The author, much to his disgust, went with the flow, and vowed never again to listen to the advice of a committee of "experts".
Theorem: Any expert can prove that something cannot be done.
Corollary: If you wait for experts to agree on something, you will wait a long time.
It is of much relevance that the Fidonet design and initiative was never put to any committee. The three of us in the Rhodes Computing Centre "went it alone", without budget, with scrounged hardware, but with a firm belief that we could do it. The same applied to the links into Africa.
Letter from Sol Lederman of SRI
Having had the Bitnet acceptance rejected by local "experts", the author had nothing to lose by writing directly to the organisation that at the time seemed to be the body that had any kind of control over the Internet, viz SRI International in Menlo Park, California, and asked various questions about connecting to the Internet. This was on 6 May 1988.
A reply was received, dated 28th July 1988, from a Sol Lederman of SRI. In essence, a connection to the Internet was unlikely to be approved, but he suggested that we investigate four other networks. These were
- TWICS (Two Way Information Communication System) serving mainly Japan but with a link to South Africa;
- FidoNet, with a reference given to Dave Pedler based in Johannesburg;
- Telenet, a commercial Public Data Network linked into South Africa's SAPONET;
- Tymnet, a commercial Public Data Network.
Of all the things that gave inspiration to the author, it was this letter. Someone out there had clearly taken considerable trouble to search out this information for us, and to give some kind of guidance. It came at a time of fairly deep depression, when it seemed that there was simply no hope of getting international email for many years, and that there were too many people in this country who simply accepted that nothing could be done about this situation.