In the foreground of this work stands the question if communication should be understood as an interpretation process or as an instructional phenomenon. The interpretative view of communication is the one advocated by Shannon's classical communication model of 1949. This (technical) model has even penetrated the human sciences and is the base of todays computer communications. It's main characteristic is the introduction of a coder/decoder pair separate from the information source and destination - a view which we call the PDU paradigm. In terms of computer communications, the coder/decoder entities are known as protocol stacks which exchange highly structured protocol data units (PDU). All a receiving protocol stack has to do is to analyse each received PDU and, according to the data values found, react by either changing an internal state, by generating a new PDU or by delivering a 'decoded' message to the destination entity. The new paradigm, called messenger paradigm, replaces the exchange of data values by the exchange of instructions. Such groups of instructions, named messengers, are no longer analysed when received: they simply have to be executed. Part 1 of this work examines the messenger paradigm from a rather philosophical viewpoint by using the conceptual framework of semiotics, but also by relating the messenger paradigm with the virus mechanism found in biology and informatics. Part 2 applies the messenger paradigm to some protocols of computer communications and introduces the key elements of a first execution environment for messengers. It is shown how Stenning's sliding window protocol can be reformulated within the messenger context, thus leading to an implementation of this protocol which does no longer require preinstalled protocol entities. Unexpected theoretical problems become manifest when one tries to relate a messenger-based protocol realisation with its related ordinary PDU-based form. One of the conclusions of this work is that there may be protocols which can be expressed in terms of messengers but which have no equivalent representation under the PDU-paradigm. Part 3 finally investigates the consequences of applying the messenger paradigm to computer communications in general. Messengers offer in principle complete liberty for the structuring of computer communications. However, in order to be useful, some common structures (i.e. conventions) are necessary. The problems of setting up a meta-communication architecture are then discussed whereas it seems that new insights may come from extending the principle of computational reflection to a distributed environment, but also from existing theories of organisation, that is, cybernetics and sociology. More research is needed in order to better understand the implications of the messenger paradigm not only for computer communications, but also for a general theory of communication.