On Tuesday 28 January 2003 06:08, Mary K. Conner wrote:
>
> I'm speaking of triple or higher checks where all residues
> agree.  The only reason to do those other than the exponents that have only
> 16 bit residues is to check for cheating.  If those kinds of checks need to
> be done, they ought to be done with intelligence, not by random poaching.
>
As someone who has done a fair number of these, and is continuing to run, I 
think I am operating responsibly. I am selecting for triple-checking those 
exponents where both (or all) the entries in lucas_v.txt were contributed by 
the same user id.

So the only way I would be "poaching" is if someone else has already 
"poached". 

Naturally the situation will sometimes arise by chance that LL test and DC on 
a particular exponent happen to be assigned to the same user, particularly 
those who process a lot of exponents. I'm also aware there are other ways of 
cheating, but this method seems likely if the idea is to boost league table 
rankings.

As it happens, I have not found any evidence of cheating, but I have exposed 
a problem which resulted in a few (very few, well, two to be exact) exponents 
being accepted as "double checked" resulting (I think) from a single run 
being reported twice.

This work is complete up to the mid-5 million range; the leading edge is just 
below 6 million. The total number of exponents involved is not enormous.

As I proceed, I'm also completing any trial factoring which might have been 
missed, and running P-1 to "high memory LL test limits". A number of 
exponents have been eliminated from lucas_v by finding a factor.

Incidentally, one of the factors I found (but only one, so far) has appeared 
in my PrimeNet personal status report. I don't know why this should be.

I have a couple of people working with me on this. If anyone else would like 
to get involved, please e-mail me. But don't expect exponents significantly 
smaller than those you might get for normal DC assignments.

I'm also working, at low priority & again with a couple of helpers, at 
completing triple-checking for all small exponents (under 1 million). Why? So 
far as I'm concerned, it's something useful for a couple of slow systems to 
do whilst they're acting as room heaters! I certainly don't regard this 
sub-project as "important", it's just that the systems I'm employing are too 
slow to be of much use to PrimeNet, even for factoring assignments.

Regards
Brian Beesley
_________________________________________________________________________
Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

Reply via email to