Coase's theorem says that when transaction costs are zero, it doesn't really matter which way we allocate the rights, and I have a fleeting feeling that it also applies to IP. Thus, from economical perspective the question becomes what arrangement minimizes the transaction costs.
I should say that I agree with this. Transactions costs are the heart of the problem - and unlike the cost of copying and distribution they aren't going away. It is true that the internet lowers the costs, for example, of micro-purchases, so that IP owners could potentially contract with lots of people, or collect small payments from many people. But while pure transmission costs are either trivially small, or will be shortly, transactions costs are going to zero. In the final analysis, there is the time needed to read and understand an agreement, and technology is helping a great deal with that part of the cost.
As Mikko says, absent transactions costs, IP wouldn't matter that much either way. That isn't an argument that we should have IP if there were no transactions costs. That is, if transactions costs were trivially small, it would be easy enough to finance new creations/inventions by agreeing to create/invent only if the beneficiaries paid in advance. But in fact the transactions costs are quite high - figuring out who the beneficiaries are, how much the product is worth to them, and negotiating agreements with them is pretty expensive. The transactions costs going the other way - when there is IP, trying to prevent people for putting stuff on P2P networks, for example, is also quite high.
