No particular quotes - lots have people have had interesting and
relevant things to say.
1) I have no problem with the likes of David Campeau (aka
diamonddave) as, whatever systems he is using, the exponents seem to
get cleared in a reasonable time (_much_ faster than the reports
indicate) - and he does appear to be acquiring exponents in a way
which is not only accessible to everyone, but which he has even gone
to the trouble of documenting. This is _not_ "poaching", which is the
quite different activity of working on exponents assigned to someone
else.
2) There is no doubt that a certain amount of "poaching" is going on.
Will the offenders (they know who they are) please desist. Poaching
often results in early triple-checks being done, at this stage this
is a waste of resources.
3) Cutting the 60-day exipry date may well have unwanted
consequences. During the last week I obtained an assignment (for a
double check in the 4 million range) which disappeared from my status
report after a couple of days. I guess whoever had the exponent
before finally returned the result - deliberate poaching does not
seem likely in this case, and it's very doubtful that anyone would
have a system powerful enough to start & finish the job in the time I
had the assignment.
If anything, we ought to be increasing the figure, as assignments
take longer to complete these days. Perhaps we could/should have
different expiry dates for different assignment types, but this would
probably involve changes to PrimeNet which aren't strictly
neccessary.
4) Please don't forget the many people using manual assignments for
one reason or another. Often non-Intel systems which can't use
PrimeNet because of the hangups over the security code. Shortening
expiry times or the maximum amount of work queued makes it more of an
effort to keep these systems contributing. (Of course, if you want an
all-Intel project...)
5) Finally, a couple of suggestions.
a) I agree with the correspondent that said that everyone should have
an equal chance of picking up recycled small exponents. A practical
way of achieving this would be to run the job which recycles
expired/returned exponents at a random time each day instead of a
fixed time. (Start the job at 0600 GMT but have it immediately sleep
for a number of seconds picked from a uniform random distribution
minimum 0, maximum 86399).
b) As a deterrent to poachers, could I suggest that any results
submitted by anyone working on exponents assigned to someone else be
credited to the "poachee" rather than the poacher.
c) I would suggest that v20 sorts assignments so that they are
executed in order of increasing exponent (once trial factoring prior
to LL testing has been completed) rather than being executed in the
order they were assigned.
Regards
Brian Beesley
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
