On 9 Mar 2001, at 2:24, Scott Kurowski wrote:
>
> I would instead recommend a broader strategy that expires exponents based
> upon the assignment age (days run) and current iteration at null (not
> started), perhaps at 60 days age. Did someone already suggest that?
I think that's a _brilliant_ idea. It's probably not hard to
implement, either.
The downside that needs to be discussed is what happens to manual
tests (e.g. using programs that don't have PrimeNet client code). At
present these can be extended (kept alive) using the manual testing
page, but there is no way of signalling the current iteration using
this form. I think PrimeNet would probably have to be made aware of
which assignments were being handled manually and have these
assignments exempted from "early release" on the grounds of apparent
inactivity. AFAIK this shouldn't destroy the policy, as those using
the PrimeNet manual testing forms are probably amongst the more
conscientious contributors. (?)
> [... snip ...]
> The result would be slightly more frequent 'opportunistic' double-check
> passes as the machine forges along to complete the then-redundant test,
> probably after the reassignment machine finishes. Maybe that's a good
> thing.
For LL test assignments, this is undoubtedly true. For double-check
assignments, we'd probably end up with a small number of "accidental"
triple-checks. IMO this is not a real problem, _provided_ the number
really is small.
>
> George Woltman manages the server's individual exponent and range
> assignments from time to time. If overriding a 'squatter' is important
> enough, he could do so manually. However, if server changes are necessary,
> we defer to him for those requirements.
>
Indeed there are frequent incidents of assignments which are
"lagging" being "pirated". Obviously widespread abuse of the
automatic assignment scheme should _not_ be encouraged, but anyone
who is really upset can actually do something about it even now -
providing they don't mind PrimeNet not crediting the CPU time!
Regards
Brian Beesley
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
