[squeak-dev] Cryptography: Fortuna>>nextInt:

Chris Muller asqueaker at gmail.com
Tue Jul 6 15:52:40 UTC 2010


Hi David, yes, absolutely, Fortuna>>#nextInt: should be deleted,
because it is a flawed implementation, as the following script
demonstrates:

"explore it"
|f|
f:=Fortuna picker.
((1 to: 1000) collect: [ : n | f nextInt: 6 ]) asBag sortedCounts

Note the consistent, heavy weighting for "1"..

The method should be deleted so it can inherit the generic one in the
superclass.

Thanks and Regards,
  Chris


On Mon, Jul 5, 2010 at 2:58 PM, C. David Shaffer <cdshaffer at acm.org> wrote:
> I missed a method in my merge of Chris' branch:
>
> Fortuna>>nextInt: was removed by cmm.7 with the comment:
>
> "Removed faulty optimization of Fortuna>>#nextInt:.  Now inherits slower
> but correct one from superclass."
>
> Chris:  It looks like the original version has persistent (with
> underscores fixed) in the latest version.  Can you confirm that I should
> remove the version below?
>
> Feedback from others?
>
> Thanks!
>
> David
>
>
>
> Fortuna>>nextInt: anInteger
>    "Answer a random integer in the interval [1, anInteger]."
>    | r high bits highestMultiple |
>    anInteger strictlyPositive ifFalse: [ self error: 'Range must be positive' ].
>    high := anInteger-1.
>    bits := high highBit.
>    "Calculate highestMultiple of anInteger, so that we can simply divide r by anInteger and use the remainder for the random value."
>    highestMultiple := (1 bitShift: bits) truncateTo: anInteger.
>    [ (r := self nextBits: bits)
>        between: 0
>        and: highestMultiple ] whileFalse.
>    ^ r\\anInteger+1
>
>
>



More information about the Squeak-dev mailing list