A message is composed on the sending Radio Telex.
Consulting some internal tables (constantly updated) the Radio Telex decides which (of possibly many) nearby Radio Telexes to send the message. If the address is unknown, it is sent to a known smarter machine.
The Radio Telex routes the messages, by radio, to a similar Radio Telex nearby, which will get it closer to its destination. That Radio Telex in turn looks at the address, and moves it closer to the destination. Finally, after many hops, and anything from seconds to days later, the destination Radio Telex recognises a packet for itself and flags it as such for the recipient.
Note that every Radio Telex can, and does, perform routing. This is a distributed switch, with no central exchange. It implies that dynamically updating routing information, of varying complexity, resides in every Radio Telex. While a message is being composed on one Radio Telex, it may be routing many other messages one hop closer to their destination.
In practice, a few nodes may be strategically placed to assist message movement - particularly on moving vehicles, such as trains, and goods trucks, for rural areas. Indeed, these may themselves become moving post offices, with varying degrees of participation by the carrier.
Note also, that the higher the density of Radio Telexes, the quicker and more redundant the routing paths become. After a year, for instance, it may not be necessary to have a Radio Telex on the train, because sufficient Radio Telexes exist down the line that it is quicker to simply hop down by radio links. However, traffic may have reached such a level that it is advantageous to keep these strategic nodes.
The alternatives are few. Cellular telephones have become popular in the West, but even there are too expensive to buy for the ordinary subscriber.
However, present cellular phones emulate and compete with the excellent service provided by landline based companies.
Research is going on in satellite-based communication. In particular, Motorola has the Iridium project, to put a constellation of 77 satellites into low-earth orbit, to be used for cellular phones. Costs Costs break down into two portions. Firstly, the cost of the Radio Telex, and secondly the cost of the service. Today, the major cost is for the service. In fact, with cellular phones, often the service providers are willing to give the subscriber the phone, in return for locking them into a service contract for a year or so.
For a third world subscriber, this puts the service out of reach. For good progress, it is necessary (as in the USA) to be able to use the service without consideration of per-call costs - or very low per-call costs. This way, it becomes an investment in the infra-structure of the country, like roads. Moreover, a considerable part of the expense in our switched networks is the accounting - who is going to pay for the call, and how much ? If we rid ourselves of this, a great deal of simplification is possible.
Much of the expense of the present networks comes from the need to provide virtually instantaneous communication, for voice, and to provide switching capacity to handle the largest potential loads.
Thus, when a fax machine wishes to send a message across the country, a real time connection is set up. When the last page has passed through the sending Radio Telex, the recipient may pick up the complete fax. However, this is not always necessary. The recipient may have been content with getting that fax in ten minutes.
Without the requirement for instantaneous communication, system performance requirements are eased considerably. Now, packet switching becomes viable. Peak loads are smoothed - packets can be re-routed past system bottlenecks.