On Monday, 4 March 2013 at 04:18:10 UTC, Rob T wrote:
> On Saturday, 2 March 2013 at 17:40:58 UTC, Joseph Rushton Wakeling wrote:
>> Hello all,
>>
>> Can anyone advise on the theoretical basis for the unpredictableSeed method in std.random?  I've tried googling around for the theory of good thread-safe seed generation methods but haven't really found anything. :-(
>>
>> Thanks & best wishes,
>>
>>    -- Joe
>
> You can use the real time clock, which should have nanosecond precision. It should be very hard to predict because the clock will fluctuate based on environmental factors. I don't know if all architectures have an adequate real time clock however if portability is needed.
>
> --rt

Maybe you can try to connect an external hardware device (e.g. arduino) and read some params from real world... :)

On 03/04/2013 09:58 AM, Andrea Fontana wrote:
> Maybe you can try to connect an external hardware device (e.g. arduino) and read
> some params from real world... :)

Yes, there are nice options here ... :-)

However, to re-focus the discussion -- I'm not so much asking "How do I ensure my own code is statistically safe?", as there are lots of ways I can go about that.  I'm concerned with the theoretical and practical justification for Phobos' existing unpredictableSeed, and possible superior alternatives that could reasonably be implemented _for Phobos_.

On Monday, 4 March 2013 at 11:04:46 UTC, Joseph Rushton Wakeling wrote:
> On 03/04/2013 09:58 AM, Andrea Fontana wrote:
>> Maybe you can try to connect an external hardware device (e.g. arduino) and read
>> some params from real world... :)
>
> Yes, there are nice options here ... :-)
>
> However, to re-focus the discussion -- I'm not so much asking "How do I ensure my own code is statistically safe?", as there are lots of ways I can go about that.  I'm concerned with the theoretical and practical justification for Phobos' existing unpredictableSeed, and possible superior alternatives that could reasonably be implemented _for Phobos_.

I found this which seems to be what Phobos duplicates
http://www.cryptosys.net/rng_algorithms_old.html

The theory appears to be no more than an ad-hoc attempt to find something unique and hard to predict across threads, processes and machines.

The superseded and improved version uses a hash of more potentially unique values
http://www.cryptosys.net/rng_algorithms.html

Clearly we're lacking a real solution, and IMO the solution should be hardware devices that come with standardized random generators.

--rt