[metapost] Re: quality of randomness

Taco Hoekwater taco at elvenkind.com
Sun Mar 13 10:12:04 CET 2005


Hi Larry and Karl,

On seed values:
--------------
Looking back, I failed to make clear in my quoted message
that I do intend to add a 'system dependancy' remark about
randomness in the new version. My point was just that it
is beyond my control to actually do something about it.

I certainly will write up and/or change something in the code
to make sure that this issue is conveyed more clearly to MP
implementors from now on.  But the actual situation may or
may not improve, regardless of any added remarks, because
aquiring seed values from the system will remain system-
dependant.


On general quality of randomness:
--------------------------------
MP currently uses the system-acquired value only to set the
seed for its internal code, from then on it uses an internal
calculation, so that it can ensure replay-ability. It is
possible (likely?) that this internal program code does not
generate numbers that are as random as you would want them
to be, and that there is room for system-independant
improvements in this part of the program.

I'd be more than happy to merge in code that implements an
improvement, assuming it can be demonstrated that the new
code is both replayable (from a specific seed value) and
portable.


On prioritisation of MP improvements:
------------------------------------
When I say something has low priority on my list, that does
not mean that something cannot be added to MP straight away,
just that I personally will not do research or coding on it
in the near future. But there is no rule anywhere that says
I am to be the sole developer ...

If someone provides an easy-to-merge patch (by this I mean that
the actual patch should follow those guidelines I posted some
time ago), I will add such a patch as soon as possible. That
is independant of the problem's ordering on my personal todo
list.

Greetings, Taco



More information about the metapost mailing list