The most fundamental protocol is merely the idea that some packets of information transmitted between autonomous systems (nodes) should have a value indication that represents transfer of a monetary obligation between the operators of those nodes. Implicit in this is that the operators may retain some of this for their own benefit as a quid pro quo for contributing their part, moving the data, to some larger activity. The net obligations are recorded at the interfaces and operators of adjacent nodes will occasionally transfer money according to those records and thus operators that provide a net service to the world will receive net income or a net expense according as they are are net contributors or net consumers of information services including data transmission. Ideas of incentives and reputation are merely theory that explains why the protocol works.

Note that the above idea does not require agreement on the size or denomination of the worth field of a packet, except pair wise agreement at the two ends of each link. The standard is thus extremely flexible in that local evolution can proceed without disrupting the global functioning of the protocol. In this it is like SLIP and PPP that operate in the same network. Standardization of packet specs is only for the convenience of not having to write too much code.

We speak here as if the worth fields are integrated into link level formats. Worth fields may, of course occur anywhere so long as the sender and receiver agree that they are worth fields and agree on their denomination. The sender and receiver must have some financial agreement. Fixed link connections are natural in that the operators of the two ends of an unswitched link necessarily already have some sort of long term relationship. I don’t know how to apply these ideas to ad-hoc radio link networks of mobile nodes. Whatever protocol carries the worth fields should protect against loss or duplication of data lest payment conflicts arise, at least for larger worth fields. If this sort of money is to pay for functions of switching and data transport then formats supporting those functions must carry worth fields. This means to me that link level protocols must support worth fields. If link level protocols support worth fields it is not clear why other protocols need do so.

That a node is able to take all the money in any packet will surprise some. The incentive not to do this is repeat business. The cost of establishing a relationship with nodes, and arranging for the physical transmission holds the node operator hostage, to a degree, to his investment. One does not steal all the money for the same reason that a bank does not. Trust must be built in order to maintain a business. It is the iterated prisoners’ dilemma.

Two rather fundamental protocols are the datagram and circuit protocols. Unless one of these is wide spread I don’t see how services such as scouts or guides can develop. A scout must contend with whatever versions of these protocols occur in the nodes it serves. This is a good reason to propose early datagram or circuit protocols. Some circuit qualities require that the nodes thru which they pass be “qualified”. Some of these qualities should be proposed early but this is an area that we believe will continue to develop far into the future. Since guide services will be necessary for most applications, early protocols to invoke guides should be established.

Name servers require early work. I think that some of the Internet name server proposals may suffice. Indeed some of the name server proposals can serve Internet and Silk net at once.