Fidonet Issues
From ZaInternetHistory
| Table of contents |
Overview
The Fidonet system played a major role in the first email gateway at Rhodes, so it is worth spending a bit of time discussing issues surrounding it.
As far as is known, Fidonet was developed by Tom Jennings out of a PC-based bulletin board system (BBS) system. This was in the 1980 era. By 1985 there were some 400 nodes in operation around the world, but mostly in the USA. BBSes were popular with computer hobbyists. In essence, a registered user could load software to and from such a system via a dial-up modem, and could leave messages for the system operator (sysop) or for other users of the same BBS system. Tom Jennings extended this concept on Fidonet, such that messages could be left for users of other BBSes, and thus a system of interconnections between BBS systems took place. Given the cost structure of phone calls in the USA, where calls within most cities were free, Fidonet was build around an hierarchical system where PCs in one city would communicate directly with a hub PC via a local (free) phone call, and the hub PCs would communicate with other hub PCs in other cities using efficient aggregation and compression techniques. By carefully planning which hubs contacted which other hubs, the cost of the long-distance calls could be kept as low as convenient, and messages flowed via a store-and-forward mechanism.
A very important aspect of Fidonet is that it allowed some bright and enthusiastic persons to get involved in networking. The efforts that these persons put into this exercise were enormous, and it was done with full self-motivation. The costs were met out of their own pockets, so there was every reason to get the last ounce out of the computers and out of the network. This led to implementing all kinds of cost-saving schemes. Eg, all files of any significant size that were exchanged across the communication links were compressed by the most efficient compression algorithms available, and there were restart points during transmissions that catered for spurious disconnections. A scheduling system was in place that allowed for cheaper time-of-day telephone rates to be used. And there was more and more of this type of thing. Not to be forgotten is the amount of help that was readily forthcoming from Fidonetters world-wide. It is very difficult to overstate the amount of help that we received in this regard from Randy Bush, who put in an incredible amount of effort to help us.
Fidonet does not scale very well. Every Fido node needed to be registered in a central file, and this central file had to be installed on every Fidonet system in the world in order to ensure reachability. Maintaining this file consumed significant effort, and consumed precious bandwidth. But beggars could not be choosers, we were very grateful for what Fidonet offered us.
Fidonet was developed at a time when the Internet was to all intents and purposes limited to the research and academic community in the USA. Before the end of the 1980's, it was apparent to all that it was highly desirable that the various email systems of the world should interconnect, by means of email gateways. Such an interconnection would, for example, allow users of the UUCP and Bitnet networks to exchange email with one another and with users of the Internet. By some means or other, and no later than by 1988, the Fidonet system could exchange email with these systems as well. Someone had developed an "add-on" package called UFGATE, which ran on the Fidonet PC and acted as a gateway to/from a host running the UUCP communication protocols. There was a large UUCP community in the USA in that era, and the UUCP network had several gateways into the Internet itself. For some reason or other, UFGATE did not appear to catch on in South Africa, and the group at Rhodes were unable to get it to work successfully.
The design of the Fidonet email addressing system, and the method of holding addressing information, was not directly compatible with the standards for email exchange between two computers as spelt out in the Internet standard known as RFC 822. Just for one thing, the Fidonet system did not support the concept of cc:ing a message to others, and this was added as a kludge (as were other features, like addressing email to someone on the Internet).
In itself, the vagaries of Fidonet were not a major problem for the team at Rhodes. For one thing, the Rhodes mailing system on the Cyber had a number of shortcomings as well. However, the standards of the Rhodes mailing system could readily be changed, those of Fidonet and RFC 822 had to be adhered to. The author drew up a specification for moving the email between the Cyber and the Fidonet system at Rhodes, This Fidonet system was called "Settler City", after a nickname for Grahamstown. (A large number of British settlers had moved to the area in 1820). The author's specification was in a document named MAIL000. Jacot and Dave did the implementation, with Jacot handling what had to be done to the Cyber in order to move email into and out of the mailer there, and Dave doing likewise on the Fidonet system. Fortunately, the experience that we all had in making the Cyber exchange email with the Bitnet-style JNET package that we purchased for the VAX in the Physics department, and in communicating with the IBM host at Potchefstroom University, stood us in very good stead.
Fidonet as a Carrier
It must be understood that we used Fidonet merely as a carrier for our email. Very few users at Rhodes used Fidonet directly. Typically, email was created by users of the Cyber (or the VAX in the Physics department), and read by users on those systems. The Cyber recognized that the recipient address in the email was not for local delivery, and so it put it in a spool file that was subsequently transferred to the Settler City Fidonet system by means of a file transfer. Similarly, email arriving into Settler City would be examined to extract email that was destined to a Rhodes computer (and eventually to any of the computers at Uninet sites), and it would be re-formatted and stored in a spool file for transfer to the Cyber.
Apart from email, there were some usenet newsgroups that were somehow brought in, as well as some Fidonet "echo" groups. There was no transfer of newsgroups between the Cyber and Fidonet, any distribution that took place had to be done to other Fidonet systems directly. This in itself was not too limiting, email was a very strong means of communication, and there were many things that could be done from a "mail-only" system. Some listserv hosts could distribute newsgroups, and there were ftpmail servers as well. We never really understood why and how the sites that ran these things did so for free, but we were extremely grateful that they did so.
Addresses within the Fidonet system, when seen from the Internet, consisted of a user-id built up of firstname.surname@{fidonet-system}. On the Cyber, the user-id's that we used at Rhodes consisted of four characters, being a hangover from the Maximop era. Eg Jacot's user-id was ccfj. The Cyber that was used by researchers at Rhodes was named RURES. The kludge used within the Fidonet system to give an email address to Jacot was then ccfj.rures@{Settler-City}. The {Settler-City} part was quite convoluted but met the standards of Fidonet.
For the record, Settler City was initially assigned the Internet-style address of f19.n490.z2.fidonet.org. Soon after June 1989, at of the whim of some Fidonetter, the address changed to f4.n494.z5.fidonet.org. The disruptive factor of this change was enormous. This taught us a good lesson, viz never rename a host to have a different Internet address unless you really have to do so, and even then don't do it.
For the record, within the Fidonet system the addresses were of the form "Firstname Surname of <zone>:<net>/<node>.<point>", for example "Victor Wilson of 5:7106/22.3". Not all addresses made use of the ".<point>" option.
The attempts at Rhodes to link to the Internet using Fidonet started in about July/August 1988. The contact with Randy Bush was from September of that year. Mail was flowing to and from Randy Bush by about 10 September 1988, but the process to set up a more formal system started with a letter to Randy Bush dated 14 September 1988. Email could be exchanged between Rhodes and the Internet by 16th December 1988. The earliest email that Mike Lawrie has in his archives that exchanged between RURES and the Internet via the Fidonet gateway is dated 15 February 1989.
It was a significant day in our lives when the Fidonet gateway was still operating in production three weeks after it was opened for general campus use at Rhodes, in February 1989. This was because we had heard that quite a few months earlier, UCT had operated a UUCP connection to UUNET in the USA, and that this link had been closed down for political reasons after three weeks. We really could not believe that we had done better than this - we had broken through the sanctions barrier on a zero budget operation without having to resort to any cloak and dagger exercises. We had also beaten the South African government of the day, which was desperately trying to control every last form of communication in and out of the country.
Linking to Botswana
By some strange quirk of fate, somewhere in Portland, Oregon was the son of a Fidonetter called John Case of Botswana. Prior to Rhodes linking Fidonet directly to Randy's system, the Fidonet routing from the USA took place via a very long-winded path that included someone in Amsterdam. It made sense, once Rhodes had a direct link with Randy, to act as the forwarding system between John Case and his son. Email records show that steps to do this were initiated at the end of January 1989. So, from the time that there was general international email on the Rhodes campus, Rhodes was acting as a gateway for email to another African country. This trend was to continue into the future, and Rhodes built up a considerable reputation in this regard.
Pat Terry and Modula-2
A major boost was given to the Fidonet linkup by Pat Terry. Pat, a professor in the Computer Science department at Rhodes, was heavily involved in drawing up the ISO standards for the programming language Modula-2. He attended one of the ISO committee meetings in August 1988, and realised how out of touch he was with the rest of the committee, who had been doing a great deal of communicating with each other by email. Randy Bush was a member of the same committee, and encouraged Pat to set up a Fidonet system himself so that he could participate more actively in the work of the committee. Pat returned from that August 1988 trip with a great deal of enthusiasm now that he could see a real use for email. He brought with him the Fidonet version known as BinkleyTerm, and a promise of a gateway link via Randy. The team in the Computing Centre then switched from the flavour of Fidonet that was currently under test, viz Fido 11, and whatever convoluted routing (via Amsterdam) that was being tested, and started working with Randy Bush.
One cannot overstate the help that we got from Randy, it was quite incredible. This was not only when the Fidonet link was being developed, it continued for very many years afterwards, through the stages of the UUCP link and the first TCP/IP link. Pat Terry's comment puts it very precisely, viz "It was an unbelievable coincidence really that of all the possible Fidonetters any of us might have met, we just happened to get a personal contact with Randy who acted as zonegate from the US to several other zones."
Zonegate Issues
It became apparent by March 1989 that the Cyber/Fidonet gateway worked in production, could handle a reasonable load of international traffic, and could be extended to handle the email of the various Uninet sites subject only to political constraints. Under such conditions, the Uninet project of the Foundation for Research Development (FRD) in South Africa was willing to fund the costs, in particular the cost of the phone calls. It was quite clear that the volume of Fidonet traffic that would flow in and out of Rhodes/Portland connection would exceed the previous Fidonet/South African traffic by many orders of magnitude. So, an offer was made to the Fidonet community for the gateway at Rhodes to become the main gateway for all Fidonet traffic in and out of Southern Africa. Thus Settler City was offered to act as the zonegate for the region, and the costs of the international traffic would not be passed on to South African Fidonetters.
For some strange reason, the Fidonet community in South Africa viewed this with considerable alarm. Clearly, something had been misunderstood, because tests had shown that the route via Amsterdam was very unreliable, apart from having to be paid by members of the Fidonet community. After a great deal of emotional email discussion, things were sorted out.
In due course, the distribution of Fidonet into Southern Africa took place via Settler City. One of the longest lasting and significant distributions was to the University of Zambia in Lusaka. This started up when the politics of South Africa started to change for the better and sanctions were crumbling - a meeting between Mark Bennett, who was an expatriate from the UK doing sterling networking at the University in Zambia, and the author, took place in Harare in November 1991. That meeting led to the establishing of this link, which ran for at least three years.
The growth of email volumes was very difficult to cope with. The Uninet project of the FRD paid for the cost of the phone calls, and provided two 386 PCs (then considered to be cutting edge), one for Fidonet and the other for development work using SCO Xenix. In the last few months of the life of Fidonet and UUCP as the main carriers of email (ie shortly before there was a leased line running TCP/IP protocols to the Internet in November 1991), the cost of the telephone calls were at least double the cost of a leased line. Telkom (http://www.telkom.co.za/) initially refused to install this leased line on the grounds that there would be "third party" traffic flowing across it, and some bureaucrat inside Telkom (http://www.telkom.co.za/) insisted that the "common interest group" recognition did not apply to the international circuit. That the research and academic community had achieved a major breakthrough to the benefit of the country meant absolutely nothing to Telkom. Somehow or other Vic Shaw sorted that one out.
Once the leased line was in place, and had connected Uninet to the Internet, there were the diehards that would not give up Fidonet. There was also an active Fidonet community in the country, and Uninet was not going to forget the favours that it had received. Somehow between Dave Wilson and Randy Bush (others??), a means was found to reverse the original transport technique and to make the Fidonet traffic use the Internet as a transport mechanism. In essence, Fidonet traffic at the one end of the Internet circuit was gathered into a file with suitable headers and moved across the Atlantic into the Fidonet PC at the other end, and then was unpacked and proceeded within Africa as Fidonet traffic. As far as Mike Lawrie was aware, this was still in operation in 1997. It was this type of cooperation and willingness to help each other that made things work, and we all felt that it would be a very sad day were this spirit to be killed off for some reason.
Fidocon '90
Given the importance of Fidonet to Rhodes University, and the interest in Fidonet among its adherents in the country, we organised a Fidonet conference at Rhodes on 8th September 1990. Guest star was Randy Bush, whose trip was sponsored by the FRD. A number of persons who were to play a role in the Internet in South Africa attended this conference, eg Victor Wilson of Telkom (http://www.telkom.co.za/). The list of delegates to this conference has not been tracked down.
The C:\JUNK Directory
By late 1989, the growth in traffic finally caught up with us, and the flow of traffic through the Fidonet system underwent a reorganisation. For whatever reason, the redesign was such that the DOS .BAT files did very little checking before taking action. In particular, there was a directory named C:\JUNK. Hidden in the depths of the gateway process, there was a .BAT file that had the command CD C:\JUNK followed by a DEL *.*.
On the evening of 7th November 1989, the author was trying to sort out some major problem with the gateway, which was jammed. In spite of the recent reorganisation, it was running out of disk space. It is not clear why Dave Wilson was not involved in this exercise, but it matters not, the author removed the C:\JUNK directory which was, by assumption, full of junk. In due course, the CD C:\JUNK command did not change directory, and so the entire contents of another directory was wiped out. This was a key directory, and so the system was unusable.
Pat Terry somehow got very involved in recovering the system. Fortunately, the disk had recently been defragmented, and most of the files could be recovered by deft use of an UNERASE command. But it was tough going, at 3am, with Pat about to leave town for a few days within the next few hours.
The lesson was clear. More people had to understand what was happening on the gateway, and it had to be ruggedised. The author prepared a document MAIL017 that tried to do this very thing. This was on 21 November 1989. This helped, but at no stage could the Fidonet gateway be considered to be totally reliable.