[SteveEisenberg]

To stir the pot:

I wonder what they do about generating data for software testing, such as to make sure the right actions take place when mailboxes get almost full. Best practice would be to create lots of test data from scratch. Easiest practice would be to copy blocks of production data to test.

If I had to guess, I'd think that in the early days of a service like gmail, they did it the easy way, but, by now, practices have improved. However, I really have no idea.

Same for the question of whether system administrators have the ability to waste lunch hours randomly looking at production emails. My guess again is -- in the early days only. The question then becomes -- how sure are we that the early days are over?

Personally, who I send mail from and to is not a privacy concern. If mining that can stop a militant attack, go for it. But the content of my mail items (have I ever emailed, or snail mailed, a password?) should be private.