Lots of talk out there about curious patterns in Democratic counties in Florida going for Bush and all the exit polls being off only in states with electronic voting. It's interesting, but doesn't hold up under closer inspection.
Here's the site where the county results in Florida are mapped out, showing the suspicious trend that in almost all small counties Bush votes outperformed Republican registration by huge percentages. Particularly in those with optical scan machines. The implication, then, is that in these small counties the machines were programmed somehow to switch every other vote to Bush. The thing is, the counties in question are almost all in northern or rural inland Florida. Most if not all of these counties voted for Bush in 2000 too, although the margins then were not as large. Moreover, the collective votes in those counties with the biggest percentage jumps in Republican votes don’t add up to the 350,000 votes that made up Bush's margin statewide in Florida. Some of those counties are so small in fact -- 3,000 votes or so -- that I’ve thought of going there and counting the paper ballots myself to see if they match up, but I have a feeling it will be close to what was reported.
And about the exit polls being off, the non-partisan peeps at Caltech and MIT came up with this systematic analysis.