[Info-vax] VSI strategy for OpenVMS

[Arne Vajhøj]

On 9/13/2021 2:42 PM, Simon Clubley wrote:
> On 2021-09-13, Arne Vajhøj <arne at vajhoej.dk> wrote:
>> On 9/13/2021 2:03 PM, chris wrote:
>>>
>>> I tend to agree with that, but a standard C library abstraction layer
>>> and header files would allow a lot of unix related (note: not
>>> necessarily linux) open source to build far more easily.
>>
>> That stuff is useful.
>>
>> Emulating reading from /dev/urandom may not be as useful.
> 
> Only because VMS doesn't have random number device drivers with a
> central entropy pool and with access to kernel-level information
> that allows it to generate better quality entropy.
> 
> In normal circumstances, if VMS itself supported random number
> generation to the same standards as on other operating systems,
> /dev/urandom would be a _very_ useful thing to emulate.
> 
> Emulating a blocking /dev/random would also be a very good thing
> to emulate if VMS had a suitable device driver for that as well.

VMS should have a way of providing such data.

But providing a pseudo to read from is *nix specific and
in many ways an outdated approach.

Application developers should use the standard libraries for their
programming language to get the data (Java SecureRandom, Python
SystemRandom etc.). Those libraries use an OS specific
way to get the data - reading from /dev/urandom on Linux,
calling CAPI or CNG on Windows, calling a future
LIB$CSPRNG on VMS.

Arne