Large sample data sets for testing

[Greg Willits]

Anyone know of some large sample data sets available for free or < $100?

"Large" = maybe 1KB per row or so, 1,000,000 - 5,000,000 records or so, 
something on the order of a couple hundred megabytes.

"Sample Data" = I don't really care what it is (text based though, no 
need for BLOB), just something to use for testing. Maybe an old news 
feed archive or something?

I'm sure I could cope with just about any common exchange format. 
Always wanted to test the performance and scaling of some setups I have 
and things I've written, but I've never located a big (to me) data set 
to play with.

-- greg willits