Berlin, April 3, 2010: One classic piece of networking history, the INWG Note 42 "Interconnection of Packet Switching Networks" which develops many of the notions that would later go into TCP/IP. To quote from page 5 on the inevitable minimality of any "super-network" virtualized from local networks using gateways implementing one common inter-network protocol:
We gradually come to the conclusion that users and gateways cannot assume anything about other networks but the simplest possible properties, the reason is that gateways cannot give to networks properties that they do not have. They can only screen out undesirable or unusable ones.
One might also think of gateways as nodes of a super-network, in which transmission lines happen to be local networks. Then, one could implement within gateways whatever properties should be deemed fit for inter-network communications, using more of less of the local features to carry messages between gateways.
Actually, this last approach boils down to the construction of yet another network, and ultimate at that. The very feasibility of such an objective can be seriously in doubt. Should network interconnection depend on a universal network, schisms would indefinitely raise the problem one level up.
This is not to say that a general inter-network agreement will never happen, but that it will happen gradually, starting with a few recommendations on fundamental issues, rather than a whole network.
Pouzin (1973) Interconnection of Packet Switching Networks (INWG Note 42) (permission to publish here granted by email from Pouzin)