Re:Googlebot’s JavaScript random() function is deterministic
by skrebbel on Wednesday February 07, 2018 @06:34AM
(#16323761)
Totally down nitpick alley yeah, but: Not if fed by an external source of randomness (eg cosmic rays).
[ Parent
| on HN ]
Re:Googlebot’s JavaScript random() function is deterministic
by philbarr on Wednesday February 07, 2018 @10:49AM
(#16325211)
> (eg cosmic rays).
YouTube copyright violations?
The chance of getting a response from Google support?
Re:Googlebot’s JavaScript random() function is deterministic
by skrebbel on Wednesday February 07, 2018 @11:17AM
(#16325434)
I love those. A bit easy to game if you have the right job at Google, but still this is cool.
I'd also assume that the Twitter firehose could be a great source of randomness.
Re:Googlebot’s JavaScript random() function is deterministic
by Godel_unicode on Wednesday February 07, 2018 @11:51AM
(#16325811)
Since we're nitpicking; deterministic effectively means same input produces same output. Input includes the seed. All PRNGs are deterministic.
Re:Googlebot’s JavaScript random() function is deterministic
by squeaky-clean on Wednesday February 07, 2018 @12:21PM
(#16326140)
Javascript does not allow you to set the seed for the default PRNG. So for our purposes it's essentially non-deterministic. You need to use a custom library if you want to be able to specify your seed.
Re:Googlebot’s JavaScript random() function is deterministic
by Godel_unicode on Wednesday February 07, 2018 @05:55PM
(#16328724)
Continuing the nitpicking: "JavaScript" isn't actually what prevents setting the seed, you're talking about a common implementation detail of specific JavaScript runtimes (Mozilla’s documentation says their seed can't be changed, for instance). A runtime developer could, as Googlebot apparently does, seed the prng with a constant value. You could also use some user-provided constant as the seed. All three are compliant with the letter of the spec.
Bottom line: the PRNG is, as all PRNGs are, deterministic. The user-facing math.random API using the underlying PRNG may or may not be. Those are distinct things.
From the spec's description of math.random with regard to random or pseudo-random: "...using an implementation‑ dependent algorithm or strategy."