Notes on Computer Communications in Developing Countries [Langer]

The focus of discussion so far has been on meeting the conventional email and news needs of a relatively small number of westernized elite developing country users, especially those connected with academic institutions or development aid organizations, who need computer communications with each other, and with correspondents in developed countries, for much the same reasons that similar sorts of people need those facilities in developed countries (but who are facing special difficulties meeting those needs in developing countries).

This focus is natural, given who is discussing the issue, and I do not deny the importance of finding solutions. But we should also consider whether computer communications may also be an "appropriate technology" of wider application in developing countries.

I will contribute to the current discussion later, and perhaps try to sum up some of the requirements already stated by others as well as adding some more along the same lines. But it seems to me we should not lose sight of the potential use of computer communications as a CHEAPER extension of, and substitute for, other communications (in addition to it's well understood role as a more expensive, but more powerful form of communications where other forms are fully developed).

We should view computer communications primarily as an aspect of communications rather than as an aspect of computing. As well as adding a communications facility to existing computer wordprocessing, as would be the most important thing in developed countries, we should also look for ways that very cheap computing equipment can be introduced to supplement other forms of communications such as radio, telephone, telegraph, telex, fax, post and publishing, in developing countries.

For example I can't think of anything that a telex or telegraph office can do, that the same office equipped with a toy computer such as a Commodore 64, with a simple dot-matrix printer and a cheap modem, could not do better and cheaper and with less skilled staff.

Yet I gather telex and telegraph bureaus are relatively more popular in developing countries than developed ones, due to the lack of cheaper alternatives (including the consequence of low telephone penetration that there is less chance of both sender and recipient of an urgent message being telephone subscribers).

Perhaps computer communications can be used to spread a cheaper and superior equivalent of telex or telegraph to the village level instead of only reaching district towns or even larger centers. After all, telegraph and telex were once the mainstays of communications in developed countries.

Although packet radio has been extensively discussed, for situations where there is no phone line, the main cost of such equipment is the Single Side Band modulation and demodulation, to which is added a computer modem (modulator/demodulator) purely to cancel the effect so as to transmit digital signals instead of voice signals!

Surely, at the data rights actually required, ordinary Frequency Shift Keyed (FSK) "CW" or at least "RTTY" would do a much cheaper job, just as effectively, with less demand for RF bandwidth (and with no need for any separate packet radio "TNC" since the computer itself could easily keep up, and certainly no need for an expensive SSB transceiver or modem). Old 300 baud FSK modems that are now being thrown away in developed countries could also be made use of.

This must especially be the case if an existing satellite or other broadcast can be used for a high volume downlink, with radio transmission only needed for the low volume uplink. (Given other needs for the satellite, receiver, only a modem, or rather a high speed FSK demodulator, need be added for data traffic).

Packet radio is especially useful for very flexible networking and for real-time interaction such as remote login to a computer, whereas we may only have urgent need for point to point links for message and file transfer. Even toy computers can do the necessary message switching on a store and forward basis per message instead of per packet, as well as taking care of the error control on a low data rate circuit. They can use simpler techniques to share bandwidth among multiple links.

Literacy is relatively widespread in developing countries compared with the overall level of development, and especially compared with telephone penetration. Computers can be used to rapidly distribute text materials (including graphics etc), which need not necessarily be viewed on a computer monitor, but can also usefully be printed as hardcopy to be displayed as posters, put on conventional bulletin boards, filed in subject folders, circulated to a reading list, delivered personally or pigeonholed for an individual recipient, photocopied, duplicated or offset printed for wider distribution. This strikes me as being at least as important as the direct use of email and news by the relatively few individuals with access to a computer and phone line.

Again, even the simplest "toy" computer with a dot-matrix printer is adequate for this, with an unskilled attendant to dispose of the output as directed by instructions that accompany it.

Also, local professionals, semi-professionals and technicians such as nurses, paramedics, doctors, teachers, agricultural extension workers may find even more benefits from computer communications than academics and development agency staff with special needs for international communications. For example an extensive local Usenet style news network could help reduce the isolation of such workers in their separate towns and villages, providing rapid access to advice and assistance from their colleagues in other locations and even internationally, as well as on- the-job educational extension, helping to build a sense of community among them and thus encouraging less over-concentration in the large cities to escape isolation.

There seems no reason why the benefits of Usenet news for unix and other computer professionals in developed countries should not have similar advantages for other occupations, who have simply been slower to catch on in developed countries, due to less contact with computers. But it may turn out that these advantages are even GREATER in developing countries, due to the problems of isolation etc.

Perhaps with Usenet style news there may even be less need for flying people around to expensive international conferences that discuss development issues, and the money currently used for that could be put to better use.

Likewise the computer access to libraries and databases that is so useful in developed countries, could be even more useful in developing countries where direct library services are so poor. It might even be worthwhile for a group of developing countries to join together to put current library materials online jointly, by optical scanning if necessary, rather than wait for developed countries to overcome the legal and social obstacles that prevent immediate establishment of a fully online library system that is already technically and economically feasible. It should certainly be a LOT cheaper than each developing country maintaining it's own national and University library system with expensive purchases from western publishers. It seems to me that only developed countries disinterest in the different copyright environment that exists in many developing countries, prevents this being one of the highest priorities for technical and educational development assistance.

At the very least, computer communications should be able to facilitate access to remote libraries, by sending requests for information to librarians, even if the result has to be mailed back as hardcopy, though even a conventional library should be able to OCR scan articles for email instead of photocopying them to post.

The article on "Networking in Africa" (Vineeta Shetty, special to "Networking World", 14 October 1991, 8:41, pp 33-35 and next edition), that Elliot Parker offered to photocopy and post, is now available by emailon request to vita@gmuvax.gmu.edu (sign of the times - one should not HAVE to photocopy and post an article any more!)

I found it very useful for understanding the infrastructure background relevant to expanding email use in developing countries.

The current pattern of telecommunications development described in the article indicates that telephone tariffs would be especially high on long distance calls and even more so on international calls. Also these tariffs are likely to rise further instead of falling as in developed countries, since there is a shortage of circuits relative to existing traffic, resulting in a much higher proportion of uncompleted calls than is accepted in networks dimensioned according to developed countries standards, and pressure to invest in additional network capacity rather than more subscriber loops. Finally, lines are of low quality as well as congested and tariff structures are arbitrary and irrational (which may provide opportunities to benefit from careful attention to routing and schedules).

None of this is immensely surprising, but I believe the implications for computer communications should be examined more closely than they have been. Let us also take into account the fact that the incomes that have to pay the high tariffs are extremely low (and with a huge gap between a relatively small westernized elite that can make any direct use of modern telecommunications at all, and the large majority who cannot). (These facts are not brought out in the article, which is addressed to western networking professionals and primarily concerned with the problems faced by large transnational companies operating in Africa- for which they are even establishing numerous national packet networks, largely useless for local needs).

A theme of the article is that developing countries need to invest more in network capacity to reduce congestion rather than concentrating on an increase in the number of subscriber loops. I would have thought that it was more important to achieve as much minimal connectivity as possible, i.e. as many subscribers as possible (including at least single long distance circuits to as many places as possible), before worrying about meeting developed countries standards for grades of service. Computer communications can greatly assist in making the maximum use of minimum connectivity, and in providing cheaper connectivity where there is currently none at all, and where there otherwise won't be any until a full voice grade phone line or radio channel can be afforded.

Assuming that sub-saharan Africa represents the most difficult developing country environment (which seems plausible to me, c.f. more developed countries in Asia and Latin America), then I believe the telecommunications problems ADD to the relative usefulness of computer communications rather than posing major difficulties for use of that "advanced" technology.

Clearly low quality, expensive and congested lines are MORE suitable for computer communications than for voice communications, when used properly. If voice communication is possible over a bad phone line, then data communication is significantly easier, and simply involves a lower data rate than normal due to error correction, instead of the great difficulties getting any intelligible message at all through on a bad voice call. Likewise, repeated dialling until one gets through is less of a problem for a computer than for a human caller.

High costs of calls on the limited capacity long distance lines increases the relative advantage of text communications over voice. When international circuits were really expensive in developed countries, the bandwidth available was used primarily for telegraph traffic with voice calls an occasional luxury. (Even radio broadcasters in Australia followed cricket matches in Britain by continuous telegraph updates, adding their own sound effects).

Given high tariffs, long distance calls would be more often used for essential messages rather than longer chats - often with the phone line merely used to transfer a written message from one side to the other by voice dictation.

Email has great benefits for getting messages through relatively cheaply, by permitting a much shorter call, and conveniently without the delay for getting through or the need to transcribe messages. Fax can, and does, have a similar advantage over voice calls, but email can, when properly implemented, be even cheaper and more convenient than fax (although that is not generally realized in developed countries either, mainly due to the early CCITT standardization of fax via ordinary PSTN lines, and lack of similar standards for PSTN email).

In developing countries email should have an even greater advantage over fax due to the higher cost of dedicating a phone line to a fax machine (whereas it is comparatively easy for email to use a non-dedicated line shared with voice calls), and the higher cost of calls. (If tariffs are currently based on a minimum of 3 minute or even 1 minute calls, it may be important to persuade telephone companies to make a revenue neutral change to charge by the second so that the full potential of email for reducing traffic can be realized.)

Of course the extremely low telephone penetration and lack of circuits between the capital cities and some major regional centers (let alone small towns and villages), severely limits the section of the population that can directly benefit from any form of telecommunications in developing countries.

However that only emphasizes the greater role for computer communications, which need not rely entirely on telecommunications networks, and can make more efficient use of the available telephone network capacity.

A country that doesn't even have a single telephone circuit to some major regional center must also have poor road transport to regional centers, with consequent high costs for postal communications and distribution of publications (although not proportionately as big a problem, since postal communications and transport of publications are relatively labor intensive and labor is cheap, compared with the more expensive capital requirements for a phone line).

Even where there is no phone line, computer communications using postal (or even carrier pigeon :-) delivery of diskette media could have significant advantages over physical transport of postal messages.

Where there is a phone service, but it is expensive to subscribe to and use, computer communications can significantly reduce the long distance phone traffic and provide a hard copy bureau service similar to, but much more convenient and cheap than telegraph, telex and fax bureaus.

In developed countries there are well advanced plans to replace most postal messages by email, either direct or for hardcopy printout from a local Post Office (with automatic enveloping to ensure privacy from the postal staff). Likewise national and international publications are often distributed electronically for local printing, although "local" means a major city in this case.

That is driven by relatively high labor costs in postal service and transport, and (expected) widespread access to email for message origination. But despite the lower labor costs, the high costs of road transport may make this approach even more viable in developing countries. After all, most postal articles are printed publications or business correspondence (mainly bills) rather than personal messages or packages, so only business book-keeping services and publishers need have email access for the approach to be viable. (Though the need to maintain a daily postal service for the remaining traffic may undermine the economies unless a viable system of scanner input and private printed output for personal messages is established so that only a less frequent postal service for packages needs to be maintained).

Local photocopying or duplication of publications would not be cheaper at the small town level in developed countries (yet), due to higher labor costs and economies of scale. But it might well be worth investigating in developing countries. A network of local offices with desktop publishing equipment such as a combined laser printer/ photocopier, could produce local editions of national publications instead of separate local publications, with the local content and local advertizing either locally added in to gaps in the national edition, or fed to the national office to be included in the local layout that is sent back for local printing only. (Although this could be done with national delivery of laid out artwork instead of desktop publishing files, the delivery delay followed by local printing would be impractical for daily newspapers, so electronic delivery followed by local printing in parallel with the national printing would be essential).

This is only an extension of newsagency type services, which have already been successfully demonstrated among student newspapers in Canada using only FidoNet style mailers (DistNet). Direct newsagency style services could be used for daily bulletins to medical personnel, teachers, agricultural technicians etc.

The above implies greater emphasis should be placed on cheap equipment and hardcopy output in analysing requirements. It also implies we need to focus more on who is doing the requiring for whom and who is to meet those requirements.

Specifically:

1. HARDWARE OPTIONS

The project that is collecting obsolete computers in the UK for educational purposes in developing countries seems to me quite likely to end up with more 8088 floppy disk computers than printers or VDUs, since the latter can often be used with more modern computers and computers with hard disks are less obsolete than those without. This project should certainly be extended to other countries as I have read of American Universities and corporations simply junking hundreds of old PCs. The junking is happening now, so this should be treated as urgent, while other aspects of the problem are still under discussion.

1.1 NO HARD DISK OR TAPE DRIVE

Whereas I would advise that a hard disk is essential for most "computing" applications, some "communications" applications are quite feasible without one, so good use can be made of obsolete floppy disk PCs and even toy computers that may be available as donations. Also one may be able to use unskilled labor for "disk swapping" to archive and unarchive information as well as to do backups without a tape drive when there is a hard disk.

This requires software designed to tighter constraints than similar software intended for developed countries where access for users without a hard disk is desirable but of lower priority. (However it would probably still make sense to develope initially for users with hard disks and worry about extending to others later.)

1.2 PRINTER ESSENTIAL

I often advise people who can't afford to spend much on a computing system to skip the printer in favour of spending more on a hard disk and on a computer that will not quickly be obsolete, since they can use other peoples' printers for hardcopy until they really need their own. For developing countries it seems to me a printer should almost always be considered essential, and one should be prepared to put up with a floppy only system and perhaps even not having a proper monitor, in order to pay for a printer.

1.2.1 A possible source of cheap printers may be the letter quality daisywheels that are being junked in favour of laser printers. These could be especially suitable for cutting old style duplicator wax stencils which may still be used in developing countries instead of more modern small offset duplicators. In particular, Diablo daisywheel printers with the 50 wire "Hytype II" interface that was used with older generation wordprocessors is useless when these are scrapped, but can be converted to a normal centronics style interface using just a shift register and a software driver. (I've done this with CP/M boxes a few years ago but I don't know whether these printers have already all been scrapped). In view of the rapidity with which laser printers are taking over, a project to stop obsolete printers being scrapped could be urgent.

1.2.2 Another source is the electric typewriters that are still being used instead of wordprocessors, despite being only marginally cheaper than a computer. Potential computer communications users who cannot afford the equipment yet should be warned to at least ensure that any electric typewriters they purchase instead should have a centronics interface, despite that costing extra. Also, it may be possible to convert some older style typewriters (there used to be solenoid kits available to convert IBM slectric golfball typewriters, though I doubt whether that would work out as appropriate now that dot matrix printers are available for USD 100 or so). Second hand electric typewriters that are suitable for conversion or that already have centronics interfaces should be coming onto the market quite cheaply as they are replaced by wordprocessors. Again, there may be some cause for urgency as they are being junked pretty rapidly.

1.3. NO DISPLAY OR LIMITED DISPLAY.

Software should be designed so it can be used even without a VDU (quite commonplace in the days when VDUs were luxuries and a tty was a teletype machine). Also software should be flexible about displaying to different column widths.

1.3.1 ALTERNATIVE DISPLAYS

Use of TV sets as 40 column or perhaps even wider displays should be considered, primarily for convenience when editing text and for control - with hardcopy relied on for reading text. This might be useful for example in a school that has a TV for educational purposes.

1.3.2. Cheap B&W TV sets can be converted to acceptable, though not ergonomically sound, 80 column monitors, through a simple bypass of the RF circuitry for direct video input. Such cheap or free B&W TVs may be readily available in countries that have more recently converted to color TV. Finally, when VGA quality color monitors are considered for non-communications related computer requirements, consideration should be given to the possibility of dual use after hours as color TV sets - either with a separate TV receiver/only unit or using the receiver in a VCR which may be available for educational purposes (and could also be used for tape backup).

2. RELATED SOFTWARE REQUIREMENTS

Apart from the software implications already mentioned above, general software requirements applicable to both developed and developing country needs, and some other developing country software requirements I will discuss later, there are some requirements that arise directly from the above.

2.1 FILE FORMATS.

For the emphasis on hardcopy above, plain ASCII is inadequate, and at least boldface and italics, and so on must be printed, even if not displayed, along with layout and graphics in some cases.

Also output must be feasible on a wide variety of printers (whether supplied as a printer-ready file to a toy computer used simply for printing, perhaps with conversion by printer drivers at a larger Message Transfer Agent elsewhere in the network, or using a local printer driver and converting from a standard wordprocessor format).

Even if no actual conversions between different wordprocessor and desktop publisher formats are provided by the network, there must at least be a labelling of "typed body parts" so that the right kind of file is used between any two User Agents, and perhaps printfile conversion for toy computer User Agents.

2.2 USER FRIENDLINESS

With the wider uses discussed above there is an even greater requirement for installation and operation to be as simple as the state of the art (plus further extensions specifically for this project) make it possible for them to be.

This is a major topic which should be discussed in detail, and to which non-technical participants can make a major contribution. It involves far more than WIMP interfaces (when there is a VDU :-) but fundamental issues like, addressing, Directory Service and so on. Even using this mailing list is a reminder that a completely artificial distinction is maintained between newsgroups and mailing lists, with no simple means for subscribing to and dropping mailing lists, unlike the situation for news groups.

User friendliness is also required for other uses of the network, such as formulating queries to be mailed to remote databases (which cannot be accessed interactively as in developed countries), and even simple file requests to mail servers. (Why shouldn't one just highlight a file that is mentioned in a message to request it, with the result that it is automatically fetched with the most efficient local cacheing system and tape transfer system to reduce international communications costs and you are simply notified when it arrives? Why should one have to formulate a mail message according to some arbitrary format, addressed to some arbitrary address, and then process the mail that comes back manually?)

2.3 NETWORK MANAGEMENT

The wider uses also highlight the critical importance of remote network management, with no skilled local operators to maintain links between leaf and relay nodes. The need to minimize costs yet maintain reliability with a wide variety of computing and communications equipment further complicates network management. Just to save costs by relaying from one town to another, or to reach a group of users in a country from outside that country, means we can't assume a model like leaf users hanging off a university computer center.

Again this is a major topic, though a significant amount of technical knowledge is needed to even appreciate it's importance, let alone contribute to solutions.

2.4 LANGUAGE FACILITIES

Transfer of wordprocessing and desktop publishing files for hardcopy can meet the most urgent requirement for use of local languages. At least initially, operators are likely to speak one of the colonial languages such as English, French or Spanish, and can use that for operations, so the usual 8 bit ISO West European character set could be adequate (whereas most current software assumes 7 bit data most suitable for English only).

However extensive local news and international exchanges between people who each speak different languages, would eventually place much greater demands on software development, which should be taken into account from the beginning.

3. ORGANIZATIONAL REQUIREMENTS

Obviously none of these nor any other requirements can be met without some kind of organization to operate the network and to develop the software (and indeed the organization) required.

It simply doesn't make sense to continue attempts at custom built software installations and specially established connections for each new user, even when dealing with relatively small numbers of users.

At the very least, some kind of "product" is required, that can be installed and used without extensive efforts for each new user, and a "service" to connect the user to other users - even if that "product" is simply self-installing configuration files for existing software normally installed by "power users" and the "service" is quite informal with no paid staff.

3.1 NETWORK OPERATIONS ORGANIZATION

Some organization has to operate the service, according to definate procedures for dealing with problems, and define requirements on behalf of users, preferably with maximum involvement from users and potential users. This organizations might also be concerned with marketing or outreach and funding etc or that could be a separate organization again - subordinate to actually operating the service really needed rather than simply "selling" what somebody just thinks MIGHT or OUGHT to be needed.

Operations may be coordinated among many different organizations in different countries, with a corresponding need to define (and organize) relations between them. It may also involve "umbrella service providers" to coordinate the activities of people who offer services through the network(s) as opposed to simply operating the network(s) that makes those services possible. Mixing those functions can also be dangerous.

3.2 SOFTWARE DEVELOPMENT ORGANIZATION

Some organization has to develope the software "product" (and define other aspects of the overall "system"). Even, or perhaps especially if it consists of volunteers, or people overworked in other jobs (including maintenance for the Network Operations organizations), it has to use professional "industry standard" techniques to define requirements, develop specifications and arrange the design, implementation and testing (not necessarily in-house). This applies just as much when heavily relying on integration of existing software. The consequences of ignoring this simple fact of life are well known to ALWAYS be disasterous.

This needs to be kept quite separate from operations, so there is a clear "handover" of finished "products" to operations and maintenance, as well as a meaningful requirements "signoff" before development starts.

In short we need to be clear who is specifying requirements on behalf of whom and to whom, or there is no hope whatever of anything useful being achieved.

3.3 DISTRIBUTED SOFTWARE DEVELOPMENT ENVIRONMENT

A simple mailing list or several lists can be used for "brainstorming" requirements and concept development etc. But if people are to be involved from different countries in both the developed and developing world, instead of the usual project team at a single location, then adequate means for communication and liaison are essential.

At the very least, mail must be used to exchange updates to formal, version controlled documents which are being worked on as tasks, according to some definate methodology (whether these documents are detailed requirements, specs, design docs or actual software). A distributed Revision Control System will be needed as well as other more usual development tools. Eventually, the work put into providing that will be useful directly to network users employing it for their own projects, as well as for developing the network itself.

Lots more, but I'll have to leave it there for now.

From: IN%"VITA@GMUVAX.BitNet"

"Volunteers in Technical Assistance"

27-OCT-1991

Subject: Article by Langer on African computers