On Thu, 18 Mar 2021 23:33:25 +0000, Farley, Peter x23353 wrote:
>Peter,
>
>I must disagree that it is not a programming interface. The problem-state
>instruction STCKE returns its value and the field is clearly defined in PoOP,
>so problem-state users deserve to know how the contents are set by the only
>program authorized to set it, the operating system.
>
>The particular application here involves generating unique application-related
>values. I can also see its application to entropy-gathering routines
>determining a unique random seed for statistical purposes.
>
>There are probably other uses I haven't imagined.
>
Uncharacteristically, I'll agree with Peter R. here (mostly). I said earlier:
A possible answer is that it's release-dependent, subject to change,
and as long as the [uniqueness] constraint is met IBM chooses not to
document it in order to retain flexibility for future releases.
My "mostly" is that IBM might provide in a software publication assurance
that the OS sets bits 112-127 to values that guarantee uniqueness with
a stated scope.
Using clock values as a source of entropy is discouraged. If a
(fe)malefactor can make a good guess at an interval during which
the clock is sampled there's little entropy available.
As for "unique application-related values" and probable "other uses",
present a business case.
But the PoOps does not disdain software topics. It contains tables of
a few dozen entries mapping TOD values to UTC. These depend on
a site's choice of TOD epoch (some still use local) and leap second
conventions. (We abandoned leap seconds for timestamp consistency
across program products.) Those tables might more properly appear
in a software publication, perhaps in the description of the TIME and
STCKCONV macros.
>-----Original Message-----
>From: Peter Relson
>Sent: Thursday, March 18, 2021 7:14 PM
>
>It should be expected that the principles of operation not have any
>information about what data is placed there, as it is specifically defined to
>be set by a program to whatever that program wants it to be.
>
>I'll bite: why would we want to document how the operating system sets this?
>It is not a programming interface.
>In what way would having this information help diagnosis (that being the only
>other reason I could think of)?
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
