About Those Needles I Spoke Of... [update]

I'm a frequentist by inclination. Rarely, not ever as I set here and attempt to recall, do I find Bayes a convincing approach to analysis. In the matter of Big Data and such, my previous musings have pointed out that if one is going to expend the effort of go needle hunting, it'd best be a really platinum needle.

Corey Chivers has done a thought experiment with the NSA data siphon. You should read it. He makes some extremely generous assumptions (generous in favour of PRISM working), and reaches a not so stellar conclusion. Just the algebra, mind, but the analysis makes clear that any actual data has to be pretty amazing for the whole thing to be worthwhile.

It's also worth noting that NSA has been doing this sort of thing for decades. When I worked in DC in the 80's, much was written about the various spook agencies. I went back a few years later, to intern with Jack Anderson (not as much fun as the current movie wrt Google, but Jack was as funny as a crutch). I even co-authored something on CIA gun running. The telephone was the beachhead. This is 1986. What NSA is doing, or is assumed to be capable of doing and won't resist, has been going on for some decades. The switch from analog telephony to digital caused a good deal of angst within Spook Nation; note the note toward the end of that facsimile about encryption and such. Back then, the spook agencies got the best mainframes IBM, CDC, and Cray could make (and no one else); it's unclear now whether that's still possible. On the other hand, I saw (but didn't keep the cite, since it seemed irrelevant at the time) the projection that 80 to 90% of traffic will soon be movies and such. More hay.

Just don't go making any sick jokes about ricin over the phone. Or, if you're in the Glass group, even thinking one.

[update]
Vinson was appointed to the U.S. District Court for the Northern District of Florida by President Ronald Reagan in 1983, and named to the FISA court in 2006 by Chief Justice John Roberts.
Here