On Thu, Feb 17, 2011 at 6:00 PM, Bob Cronin <[email protected]> wrote:

> Well, there's 8 VM's running the code and they're beat-upon constantly,
> nearly 24x7, so the probability of a fair number of them invoking it in the
> same microsecond is nonzero, I think ...

Uh, we'd be talking about them starting up at the same time and get to
invoke the "random" stage exactly at the same time...

We *are* talking about a random stage that sits there for the duration
of your application and outputs a new record when you need one? Or at
least the PIPE remains active forever (assuming the different
invocations of "random" will share the seed).
But I'm pretty sure that the Piper picked the portion of the TOD smart
enough to achieve maximum entropy, so your even your first random
number should be fairly good. I suspect you have a leak somewhere :-)

Getting a random number is obviously not a trivial issue. That's why
some applications take user input (eg mouse movement) as a seed. And
that's why we got hardware to do this (which builds on noise generated
in electronic circuits etc - expect your random numbers to go up this
night!)

| Rob

Reply via email to