[comp.society.development] Network Management

cmf851@anu.oz.au (Albert Langer) (05/30/91)

It seems to me that the most important problem for expanding network access in
developing countries is essentially the same problem as for developing
countries, namely the requirements for skilled unix system administrators
or BBS sysops to keep network connections functioning.

In developed countries that tends to restrict network access to a
relatively small section of the population who are attached to a
university or a company involved in research, which provides full-time
sysadmin staff who can handle network administration as part of their
job, or to "computer hobbyists" interested enough in computer communications
to be willing to queue up for access through BBSes that depend on many
hours of work performed by volunteer "sysops".

That is a sufficiently large section of the population for computer
networking to be viable and growing rapidly, but there is still a 
big barrier to participation even by the majority of people who use
computers for wordprocessing and could afford modems.

In developing countries there are other major problems such as:

1) Underdevelopment implies a much smaller section of the population
with access to telephones, let alone computers.

2) The phone system is unreliable and low quality.

3) Various language problems and cultural factors.

4) International communications costs, since international
connectivity is a major requirement for achieving a "critical mass"
or "take off" point given the other problems.

Lack of system administrators or BBS sysops in developing countries
seems to me a more important problem than the others because the
other problems are either easily dealt with or else cannot be dealt
with at all within the context of this discussion.

For example improved international network access may contribute
towards eliminating underdevelopment and enabling a much larger
section of the population to have access to telephones and computers,
but there is no point in attempting to eliminate these characteristic
features of underdevelopment as part of a program for providing
network access rather than as part of a much wider program in which
network access itself plays a small role.

On the other hand, the unreliable and low quality phone system
simply requires the adoption of appropriate protocols for getting
as much data through as possible. The FidoNet protocols are superior
from this point of view compared with the UUCP protocols because they
include a capacity to resume a file transfer where it was left off
from a lost connection (among others things). No doubt they could
be improved further, e.g.:

1. To monitor line conditions adaptively and abort and resume connections
dynamically in a more optimum way.

2. To configure a range of modems automatically and perhaps even
adaptively according to local line conditions and the destination
of calls WITHOUT INTERVENTION (by a skilled "modem communications
expert"). Or at least to simplify this as much as possible.

3. To squeeze more thruput by using more efficient data-compression
schemes specifically adapted for email and news text using dictionary lookup
methods instead of the present "general purpose" LZW style compression.

These are narrow technical problems which need to be addressed (elsewhere)
but which once dealt with result in the problems of developing countries
internal and international communications problems reducing to simply a
"(manageably) higher cost of network access" rather than a qualitatively
different situation from that in developed countries.

The problem of not having skilled system administrators and sysops
readily available is inherent in underdevelopment. But unlike the
problems of low telephone and computer penetration it can be 
quickly dealt with by eliminating or drastically reducing the
amount of skilled labor required. (Telephone and computer costs
are also dropping rapidly as a result of technical progress, but
I am arguing that network management requirements could drop
much faster with a small effort devoted to achieving that result).

Since networking was initially developed in academic and research
circles on the one hand, and computer hobbyist circles on the other
hand, it is hardly surprising that the technology is optimized for
the conditions that prevail in those circles - which includes
ready availability of sysadmins and sysops. 

User interfaces also tend to be unfriendly for similar reasons 
but that problem is less of a barrier to developing countries access 
(given that only an elite with telephone and computer access can benefit 
directly anyway). In any case more friendly user interfaces are 
already being developed as a result of pressures for this within
developed countries and special efforts are less essential to ensure
this happens for the benefit of developing country access.

Networking has spread into the corporate world with thousands
of PCs within large organizations being linked into enterprise
LANs and WANs. As a result of this and the development of
email connectivity generally there has been pressure for
international standards such as X.400 for email and news,
X.500 for related Directory services and both SNMP and CMIP
for network management.

I believe these standards can provide a basis for building
reliable and efficient networks that require very little
skilled labor for network management or administration. That
is a commercial necessity in developed countries today,
since corporations cannot afford to employ sysadmins in the
ratios required by previous generations of software any
more than developing countries can do so (and of course
BBS sysops won't perform voluntary labor for corporate
networks).

Unfortunately the main orientation in the corporate world
is towards WANs connected by leased lines rather than
dial-up PSTN connections like UUCP and FidoNet. This tends
to bias the management protocols towards assuming 
continuously available network connections rather than
an intermittant "store and forward" situation.

Nevertheless I believe they can be taken as a basis
for automating network management. More important, X.400
and to a lesser extent X.500 are quite well oriented
towards "store and forward" situations and can certainly
provide a framework of RELIABLE email and news that
greatly cuts down on sysadmin and sysop workloads.

I am working on a project for implementing much wider
access to networking in developing countries through
adaptation of the "freely available" software
implementing the most advanced standards. I believe
this project will also be relevant to developing
country needs and hope to keep in touch with others
here. Unfortunately I expect to be inaccessible for 
a few weeks soon, but I just wanted to say "hello"
(verbosely :-) and that I look forward to future
involvement with this group.

--
Opinions disclaimed (Authoritative answer from opinion server)
Header reply address wrong. Use cmf851@csc2.anu.edu.au