This is August 2, 1996 and I have just become aware of the ideas called “Active Networks”. Here is an excellent survey of active network ideas. I contrast here ideas from the Digital Silk Road (DSR) paper, with those of Active Networks. The ideas need each other I think.

Early papers on active networks tended to avoid incentive issues. They joined the prevailing assumption that peer pressure will counter resource abuse. One of the results of this, I think, is the suggestion that application state in nodes will be soft and will be sacrificed upon space exhaustion, much as packets are now dropped upon congestion. Another benefit of soft state is that there is less burden for the application to clean up after itself. Application clean-up may be more complex than application setup and tear-down bugs are harder to find. It is difficult, however, for an application to be continuously and ubiquitously prepared to repair damage to its distributed structure. The survey paper mentioned above clearly identifies resource allocation as a problem that requires new solutions.

In some unpublished notes DSR presumes that network resources are goods that can be owned, transferred, traded, etc. It follows that DSR money can be used to purchase network capacity. If all else fails, forgotten application components will be evicted when they are no longer able to pay their rent. Meanwhile the node operator will enjoy revenue necessary to support the dead state. The incentive is for the application builder and operator to tear-down at the end.

These are some papers that describe how a money economy can organize resources within a computer system even when no human attention is devoted to the operation of that economy. The invisible hand works for computer resource allocation. In the current DSR thinking we assume some initially simple markets where resource and resource futures can be bought to serve end users of network capacity.

The Agorics papers above describe in some detail how goods are owned, traded and employed in computer systems but I will give a bit of detail here. For a given type of good there is a widely trusted object that deals in titles to that good. There are polymorphic protocols to validate, trade, split, join and transfer exclusive rights to titles to a given good. Imagine a title for some amount of bandwidth on a given link for a specific future duration. The bandwidth office is trusted to perform the operations on such titles. When it comes time to employ the bandwidth the title can be presented to bandwidth office which will return a more immediate capability for the bandwidth. The immediate capability is probably not rescindable, transferable, etc. It is used while invoking communication primitives to claim bandwidth. The same pattern can control allocation of RAM space, fast buffers, cpu time, crypto hardware time, etc. Humans can enter this market to acquire the elements necessary for some upcoming conference just as a conference sponsor must acquire hotel conference rooms. Eventually such acquisition can to some degree be automated. Speculators in futures will presumably help both the providers and consumers of network resources hedge their bets and collectively plan new network provisions in time to support future demand. See these similar ideas.

The two ideas: DSR and Active Networks, are each radical. Doing both at once may result in too much change and too little order. Yet I think that the two ideas are powerful in combination. Indeed each may be impossible without each other.

See Integrity without Secrecy for an elaboration of these ideas.