I can imagine that bandwidth in the fibersphere for text transmission will be too cheap to meter, which means that the cost of metering would more than the marginal revenue.
[re: overload]
Rather, by the _forwardings_ of other masses of stuff, written by others. "MAKE.MONEY.FAST" is but the most recent example. Not to mention images, coredumps, etc.
I only talked about text transmission, not about arbitrary bit transmission. The situation for automatic bit sources is not the same.
I'll go out on a limb and speculate that cheap delivery makes a fee schedule of some sort _more important_, not less important.
Look, there is a cost to using the price mechanism. When the cost of the thing being purchased becomes too small, it's no longer economical to price it. That doesn't mean that it's free. It means there are other structures for accounting. One transaction per packet will almost always be more overhead than it's worth. There are other ways of paying for service, though, by connection, by total bandwidth, by link. The structure of the transaction is different, because a different thing is being purchased. Flat rate local phone calling is common. The expensive part of using a local phone switch is the switching, not the connection. Maintaining the connection is cheap.
Of course, this is up to the service providers; anyone who wishes to provded a free bandwidth link should be free to do so!)
This is irrelevant. The Libertarian-PC police aren't around, last I looked. Tim made the statement that pay-as-you-go was the obvious choice. That's not at all obvious. The accounting mechanisms are but one aspect of the transaction costs involved. It is quite possible that the only economically viable communications services are aggregated services. Whenever you have aggregation, there is some persistence, and that yields an identity. (It need not be a personal identity.) There are some interesting questions here. What is the characteristic length of that persistence? It will vary depending on the cost to do another transaction. The length of persistence is the length of exposure of an identity. What are the forseeable tradeoffs between link costs, switching, and general-purpose computing? This gives some idea about where the bounds of accounting will fall. Analyses which disregard transaction costs are unrealistic. The question is not one of paying for service; let's bury this libertarian hype against socialism right now. The question is what the structure of the communications market, both buyers and sellers, will be. I want a system with low transaction costs, because that lowers the characteristic persistence time of a communications transaction, and the smaller the time, the better the privacy. That means we have to lower the transaction costs. Let's take remailers as an example. One current suggestion is to add some sort of money system to the remailers as a condition of use. This is exactly the wrong priority at the current time. The remailers are already hard enough to use, and adding a payment system on top of that will make them used even less. Making a system harder to use increases the transaction cost. The current priorities should be to lower these costs. When the remailer system begins to be overloaded, then adding some restriction on use, perhaps by means of payment or a payment analogue, will be warranted, because it will lower overall transaction costs, trading off ease of use for throughput and reliability. What are some of these costs that should be lowered? -- Finding out that remailers exist and what they do. -- Finding a remailer to use. -- Deciding what remailer to use. -- Figuring out how to use a particular remailer. -- Formatting a message for a remailer. -- Receiving mail through a remailer. There much more need for improving the ease of use of remailers than for paying for them. The less expensive privacy is, the more privacy there will be. Privacy has non-linear benefit; the more that people are private, the better any individual's privacy actually is. Eric