Politness Costs Nothing but Can Pay Dividends

[KrazyHorse]

It wouldn't be a fair fight. I am a Monte Carlo expert

[Kuciwalker]

He's referring to the fact that the dead simplest way to generate random numbers in any programming language is going to be some variant of the following pseudocode:

rng = new random(now());
...
x = rng.next();
...
y = rng.next();
...

etc.

(and, additionally, that some people may be so lazy as to just always say "x = new random(now()).next();")

[KrazyHorse]

I recently spotted and fixed a RNG bug buried deep in our codebase that had been causing intermittent failures for years.

That may be the most impressively technical thing I've done since I've been there.

[KrazyHorse]

He's referring to the fact that the dead simplest way to generate random numbers in any programming language is going to be some variant of the following pseudocode:

rng = new random(now());
...
x = rng.next();
...
y = rng.next();
...

etc.

(and, additionally, that some people may be so lazy as to just always say "x = new random(now()).next();")

It is quite possible that somebody did the latter, but there is no reason to assume that they did. Also, the former appears to be what I was saying...

[Asher]

It wouldn't be a fair fight. I am a Monte Carlo expert

I don't care about the algos behind them. Let some masochist figure that out, the details are inconsequential to me. It doesn't matter what kind of fancy **** you're doing behind the scenes when the Indian programmer who designed those things just piped it the time as a seed.

[Asher]

It is quite possible that somebody did the latter, but there is no reason to assume that they did. Also, the former appears to be what I was saying...

If you meant the former, your question made no sense in the context of any quasi-sensible app design.

It's highly likely that each number picked for a single user is from that stream, but I would think a new instance would be generated each time the program is told to enter a new set of numbers...

[KrazyHorse]

Again, you CAN pipe in the time as the seed. That DOESN'T MEAN that the ****ing thing declares a new stream every time somebody asks for a number.

[KrazyHorse]

If you meant the former, your question made no sense in the context of any quasi-sensible app design.

It's highly likely that each number picked for a single user is from that stream, but I would think a new instance would be generated each time the program is told to enter a new set of numbers...

a) So now you know what a stream is?
b) I have no idea why you think this is necessarily the case...

[Asher]

Again, you CAN pipe in the time as the seed. That DOESN'T MEAN that the ****ing thing declares a new stream every time somebody asks for a number.

The hell are you talking about?

It won't declare a new stream every time someone asks for a number using "next" or something equivalent. I'm assuming they do a "new rand(time)" for each new user. And if they're not doing that, they're dumber than the dumbest Indian programmer I've ever met.

[KrazyHorse]

I take it back. You still don't know what a stream is.

[Asher]

a) So now you know what a stream is?

Of course I know what a stream is. I do understand RNGs and have taken cryptography courses where they play an integral role...

I just couldn't put that in context for what you were saying. It would not make any sense to me to not re-seed the RNG for each user, so it never even crossed my mind that there'd be some singleton RNG seeded god-knows-when and then constantly reusing numbers off the stream.

b) I have no idea why you think this is necessarily the case...

It's the most obvious (and probably optimal) solution to the problem. Otherwise every user in the store is working off the same seed, which is not very desirable...

[Asher]

I take it back. You still don't know what a stream is.

What's far more likely is you're not using the correct terminology to describe what you're thinking.

The stream from an API/programmer's POV is the set of numbers you can retrieve from a RNG after it's seeded.

Considering it's extremely unlikely the lotto companies wrote their own RNG lib, it makes no sense to dive into the number theory behind the libraries.

[KrazyHorse]

It's the most obvious (and probably optimal) solution to the problem. Otherwise every user in the store is working off the same seed, which is not very desirable...

Good ****ing god, Asher. Why the **** is "working off the same seed" not ****ing desirable? That is, in fact, the most desirable solution...

Also, "next" doesn't ****ing declare a new stream. new rand does that.

[Asher]

Good ****ing god, Asher. Why the **** is "working off the same seed" not ****ing desirable? That is, in fact, the most desirable solution...

You can't be serious? The point is to be random, not predictable.

Also, "next" doesn't ****ing declare a new stream. new rand does that.

That's what I just said.

[Kuciwalker]

Asher, from a pure crypto perspective using the same seed is the correct solution.

From a practical implementation perspective, inasmuch as this is vulnerable to any kind of attack re-seeding should make it more secure.