An opportunity to End Software Patents: ESP briefs Court in Bilski case rehearing

Er, as far as I was aware you can't construct and test physical machinery, 
or (say) some new drug for nothing. Building machinery costs money. 
Running clinical trials is vastly expensive. In both cases, the cost of 
<i>inventing</i> the patented entity is vastly higher than the cost of 
<i>implementing an instance</i> of that entity, given the patent.

Coming up with some patentable algorithm appears (from the quality of the 
algorithms already patented) to be a matter of a couple of hours' thought, 
if that. No significant expenditure is required above and beyond that 
which is needed to get a working instance of the algorithm. Access to the 
patent doesn't help you at all. (Note how software patents are essentially 
never used by anyone in the field for any purpose whatsoever other than as 
landmines: the system is essentially worthless as a repository of 
algorithmic knowledge. Indeed most companies recommend against or actively 
ban the use of patent searches for any purpose, because all it does is 
bring in the possibility of a wilful infringement suit, and there's no 
real chance of finding anything useful and comprehensible in there faster 
than you could think it up yourself.)