Tuesday, November 21, 2006

Underappreciated (ii)

The potential for the application of discrete mathematics beyond computer science: in biology, economics, sociology, etc.

Hal Finney's Reusable Proofs of Work is important for at least two reasons: as a good stab at implementing bit gold, and for its use of the important idea of transparent servers: using remote attestation to verify the code running on servers.

The role of Stockholm syndrome in politics. Left as an exercise for the student. :-)

Three "silent" movies which I've had the good fortune to see on the big screen with live music: The Merry Widow, The Phantom of the Opera, and Safety Last.

2 comments:

Anonymous said...

Proof-of-work does not work when people have widely differing computing capacities at their disposal. Since some of the world's largest distributed computing clusters are under the control of criminals, any cash based on how many gigaflops you have is doomed to catastrophic fraud before it has even started.

Nick Szabo said...

For applications like spam resistance using regular (not reusable) POW, your argument has some weight: to the extent these schemes assume spammers will exhaust their computing resources before end users, they may prove to be disillusioned.

But with respect to bit gold, you could make the same argument about gold and gold mines, and it would also be wrong. There are substantial economies of scale in gold mining and a handful of corporations are responsible for most of the world's gold mine output. That doesn't prevent a very competitive market in gold. (But see my comment on custom hardware below -- that is a scenario in which your argument has more weight).

For both gold and bit gold the price is determined by supply and demand. For both Hal's and my scheme the supply is determined by the scarcity at any given time. In Hal's scheme this is because tokens are turned over sequentially every time they are used and in my scheme because they are securely time-stamped. In my scheme later times will usually reflect lower computation costs and thus trade for a lower price, while in Hal's scheme the difficulty of the puzzle is gradually increased.

The worst problem, that I indicate in my article, is that somebody could come up with custom hardware with orders-of-magnitude advantage and possibly gain a near monopoly over the "mining" market until soembody else figures out a way to duplicate the hardware. This might or not beat the idea -- the market for time-stamped POW bits may adjust properly to even such a radical cost reduction -- but it's an important problem.