Eliot Spitzer briefly lays out the case for a tax on financial transactions. I think of this as basically another Pigovian tax. High-speed trading destabilizes the markets while doing little to nothing to improve the performance of those markets. A small financial-transaction tax would do very little to discourage long-term trading which is beneficial in helping capital find its most productive use, but should provide a big disincentive for short-term trades.
One question I have is whether this will end up generating as much revenue as some people claim. If it is successful in changing incentives, then the trades disappear and the tax revenue never materializes.