Replies

I'm glad I'm not the only one doing that math. I was just calculating phone calls in the US and assuming 4 calls per day and came up with less than one hard disk a month. Now, that will get boosted considerably if you include as metadata full tracking of the location of every cell phone (based on either GPS or cell tower power and direction).

Actually, adding location data wouldn't boost the storage requirements all that much. If we allow eight bytes each for latitude and longitude (three digits, notwithstanding the fact that only logitude ever needs the third one, plus four more after the decimal point, plus a byte for storing the "N-S" or "E-W" flag, which is even more bloated than allocating a full byte per decimal digit as I've been doing), that just pumps up the requirement per call from 64 to 80 bytes -- increase the above storage requirements by 25% and it's covered.

The ultimate point, of course, is that they're clearly sweeping up call content (which does use quite a bit of storage even with efficient compression), not just "metadata", en masse.

Addendum: That amount of location data is sufficient to specify any point on earth to within about 100 meters. If you need to pin it down to 1 meter, add four more bytes (two more significant digits for each coordinate, with the same inefficient one-decimal-digit-per-byte coding I’ve been using to err on the high side) and increase the figures by 35% instead of 25%.

The big increase I was talking about was not storing locaton only at the beginning and end of a phone call, but every time the cell phone pings a tower. I don't know how often that happens, but I expect that it is at least once a minute just to keep the phone company informed of where to route an incoming call. The NSA would probably count tracking the location of every cell phone in the country 24 hours a day as "mere metadata". 6 bytes will store your location to a ten foot accuracy. Do that once a minute for 200 million cell phones will result in 1.7 terabytes a day.

And all of that can be waved off as just metadata.