TCP.IP on Uninet
From ZaInternetHistory
- The title of this article is incorrect due to technical limitations. The correct title is TCP/IP on Uninet.
| Table of contents |
Overview
There can be little doubt that the TCP/IP protocols were being used internally at a number of Uninet sites in the late 1980's. Rhodes was running SCO Xenix before 3rd October 1989. Running the TCP/IP protocols at a Uninet site was a pre-requisite for any attempt to link Uninet sites into a small internet.
This section describes some of the experiences in getting TCP/IP working between Uninet sites. What is surprising is that Uninet was cleared to connect .ZA to the Internet by 25th November 1990, yet it took a further 12 months before a circuit was installed and the connection established. The roll-out of TCP/IP at the South African universities was not done with any alacrity at all, and even today (1997) there is not a good depth of knowledge of TCP/IP at all of the Uninet sites.
Rhodes-UCT Linkup
The first link that took place was between Rhodes and UCT. Vic Shaw arranged that the Uninet multiplexors were reconfigured in a way that provided a 9,600 bps async link between the two sites. Routers were to all intents and purposes unobtaineable, but we managed to purchase a package developed by Wollongong called Win/Route. This simply did not work, and the setback that this created nearly killed off further attempts. This must have been in about May 1990.
PCRoute
Given the failure of Win/Route, the attempt to establish a linkup between Rhodes and UCT almost died. Someone (?Chris Pinkham? Of UCT) came up with the idea of trying to use a public domain product called PCRoute. (The earliest reference to PCRoute that the author has is in a directory listing that he obtained from a USA site on 3 Dec 1989).
It made sense to get the networkers who were closely involved in this linkup to speak to each other face to face. Jacot and John Stevens (of Rhodes) attended a networking meeting at the University of Stellenbosch 4th and 5th June 1990, and spent some time at UCT. This had an immediate benefit, and work was in progress to use PCRoute on the Rhodes/UCT link by the 7th of June 1990. It took only a few more days to get TCP/IP traffic to flow between Rhodes and UCT.
Later that month (June 1990) a similar link had been established between Rhodes and the University of Natal Durban, with the technical work there being done by Alan Barrett.
Expansion of this type of networking was surprisingly slow. The next link in the country was from the Rhodes gateway to Potch. This was some time after the TCP/IP workshop that was held at Rhodes on 6th to 8th August 1990. The oldest SMTP mail from Potch that the author has in his archives is from Jan van Rooy and is dated 20th August 1990, but there is a message implying quite clearly that the dial-up modem was no longer in use on the 27th June 1989 - possibly because the email was flowing across one of the Uninet virtual circuits as RSCS traffic, it was most certainly not as SMTP traffic.
From then on, the roll-out to other sites speeded up, as the method of getting IBM hosts onto the TCP/IP network had been established.
Those !@$( Default Routes
PCRoute used the RIP protocol for keeping track of routes to the various networks of Uninet, and in due course for the link to the USA as well. In essence, there had to be but one default route (ie route to network 0.0.0.0) announced on but one router within Uninet, and this had to be done by the router at Rhodes, because that was where the router to the USA was located.
Again and again what would happen, particularly when an internal network was added at a Uninet site, was that a default route would get set into the configuration file of the PCRoute that provided the connection. This would cause tremendous problems, as it would cause a failure at other points on Uninet when that incorrect default route propagated across Uninet. This would happen at sites that did not have a great deal of TCP/IP experience, and at sites that mistakenly thought that they had experts and this whole routing issue was a pushover. One highly undesirable side-effect of this is that failures of international routing were inevitably and unfairly blamed on the networking staff at Rhodes.
Interop '90
8th-12th October 1990, San Jose.
Visit to SRI
(18th or 19th) October 1990 - Sue Kirkpatrick (nee Romano) and Doug MacGowan. Led directly to the registration of the .ZA domain a month later.