Status of PrimeNet

[S485122]

On the GIMPS Status page ( http://mersenne.org/status.htm) it is stated "March 1, 2006: All exponents below 17 000 000 tested at least once."

On the PrimeNet Status page (http://www.mersenne.org/primenet/status.shtml) one can see that a lot of numbers under 17 000 000 are still out for the first check. For instance for the range 14700000 14799999 there are still 79 number out for first LL testing.

Or is my interpretation of the status page wrong ?

[cheesehead]

Originally Posted by Jacob Visser

Or is my interpretation of the status page wrong ?

Not in the way I suspect you think it might be wrong. What you say about each page is correct as far as you go.

But, alas, the world of GIMPS is more complicated than either status page explains. (Furthermore, neither PrimeNet status reporting nor GIMPS status reporting is perfect.)

Example: Some people are assigned LL tests by PrimeNet, but fail to complete them, or fail to report their completed tests, in a timely fashion (and so their assignments "time out" and are re-assigned to someone else). If such a person later reports a completed result, PrimeNet will accept it on behalf of GIMPS even though PrimeNet may have that exponent recorded as assigned only to someone else (who has not yet completed their (re-)assignment even though they are submitting progress reports in a timely fashion). So we can legitimately have both a completed LL result for an exponent (correctly included on the GIMPS status page) and simultaneously show that exponent as an "out" assignment on the PrimeNet status report.

Now, I think PrimeNet is supposed to shift the "out" (re-)assignment from first-time to DC when it receives that belated completion report from the timed-out first assignee, but what if PrimeNet fails to do that for some reason, or what if the first submitted completion report is from someone to whom PrimeNet never assigned the exponent in the first place? (I.e., a "poacher" reports a result before the assignee finishes the test.) Again, we might have both a PrimeNet "out" assignment shown on the report even though GIMPS has a completed LL result reported for that exponent.

Some first-time LL reports are submitted by participants who have such a high error count on their submitted results that their exponents are soon re-assigned for first-time LL.

In all of the above cases, imagine that the assignee who currently has the "out" assignment is running the LL on a slow system. This can explain why a 14xxxxxx exponent can still be an "out" first-time test years after it was first assigned by PrimeNet.