Issue 27293: Summarize issues related to urandom, getrandom etc in secrets documentation (original) (raw)

Created on 2016-06-11 10:30 by steven.daprano, last changed 2022-04-11 14:58 by admin. This issue is now closed.

Messages (5)
msg268206 - (view)	Author: Steven D'Aprano (steven.daprano) *	Date: 2016-06-11 10:30
Write some documentation for the ``secrets`` module summarizing the issues relating to /dev/[u]random, getrandom, etc. There's a lot of confusion about these issues, and the web contains a lot of misinformation, so being able to point to the secrets docs for a summary will be useful for further discussions, and to help programmers pick the right solution. I know the Python docs cannot be the definitive source of information about OS features, but they can provide enough of a summary to allow users to make informed decisions.
msg268208 - (view)	Author: Christian Heimes (christian.heimes) *	Date: 2016-06-11 10:45
Thx Steven. tl;dr The OS' / Kernel's CSPRNG is safe-to-use as long as one uses the correct API: getrandom(flags=0) on Linux, getentropy() on BSD, CryptGenRandom() on Windows. Myths about Linux's urandom: http://www.2uo.de/myths-about-urandom/ Example why user-space CSPRNG are bad: https://gist.github.com/tiran/a9ba8c51cc7d1b75d3bc1d3f24411b4c
msg268240 - (view)	Author: Larry Hastings (larry) *	Date: 2016-06-11 18:18
As with #27292, I'm going to nosy Georg Brandl about this so he can guide us in how to approach it. My hunch is, it'd be best if we avoided specifics, and talked instead in generalities. Perhaps all that's really necessary is to consistently assure the user that the secrets module represents best practices in Python, now and in the future.
msg268241 - (view)	Author: Larry Hastings (larry) *	Date: 2016-06-11 18:19
Oops, sorry, forgot to actually nosy Georg. D'oh!
msg274713 - (view)	Author: Alyssa Coghlan (ncoghlan) *	Date: 2016-09-07 02:37
PEP 524 has been implemented for 3.6b1 in #27776, so os.urandom() itself will now do the right thing for cryptographic use cases on Linux. Accordingly, marking this as out of date - with os.urandom() and the secrets module both implicitly doing the right thing, the vagaries of operating system cryptographic RNG access can be left as obscure arcana of interest only to language runtime developers, rather than regular Python users :)