Manual reporting of non-Prime95 results to V5

[Graff]

As a test, I've tried reporting some LMH trail-factoring results (that were
generated using an OpenVMS version of Peter Montgomery's mfactor) to
the V5 server using the manual reporting form.

While the reporting of exponents with factors seems to work (in that my
V5 account gets credit for the result and the factors appear in
result queries), the reporting of new upper limits for exponents
without factors does not. Specifically, I get the error:

Error: Missing checksum. Correct the problem or email results to woltman@alum.mit.edu.

The checksum can apparently only be generated by trusted software
(and the security.c/.h files are missing from the V24.14 source
distribution), so will non-Prime95 LMHers only method of reporting new
upper limits continue to be e-mail to GW? I send him results for ~ 5000
exponents (currently taking exponents in the range 83M-84M up to
2^63) every 5 or 6 days, and the results seem to get included into the
V5 server quite quickly, but my account doesn't receive any credit for
them :-( Clearly, it would be preferable to able to report the results
directly to the V5 server.

[ET_]

DId you try using version 25.6 for factoring work?

Luigi

[Prime95]

Originally Posted by Graff

The checksum can apparently only be generated by trusted software, so will non-Prime95 LMHers only method of reporting new
upper limits continue to be e-mail to GW?

I can add your v5 user id to the list of trusted user ids. Email me your v5 user id. Most LMHers use prime95 and won't have your problem.

The current scheme is to protect the database from a malicious user sending in bogus factoring limits just to climb the stats rankings.

[Graff]

Originally Posted by Prime95

I can add your v5 user id to the list of trusted user ids. Email me your v5 user id. Most LMHers use prime95 and won't have your problem.

The current scheme is to protect the database from a malicious user sending in bogus factoring limits just to climb the stats rankings.

Will send v5 user id by e-mail.

[Graff]

Originally Posted by ET_

DId you try using version 25.6 for factoring work?

Luigi

No. My PCs use Prime95 and get work (both trial factoring and
Lucas-Lehmer tests) from the server. My LMH searches take place
on non-PCs and are restricted to trial factoring. This division of
work makes sense for my setup.

[cheesehead]

Originally Posted by Prime95

The current scheme is to protect the database from a malicious user sending in bogus factoring limits just to climb the stats rankings.

That's still a problem.

I nonmaliciously (but suspecting what might happen) entered some P-1 results on the V5 manual results page. The results were in the form

UID: ___/___, Mnnnnnnnn completed P-1, B1=12345678, B2=100000000, Wc1: ________

UID: ___/___, Mnnnnnnnn completed P-1, B1=12345678, B2=200000000, Wc1: ________

UID: ___/___, Mnnnnnnnn completed P-1, B1=12345678, B2=300000000, Wc1: ________

...

because that's how I'd run them (with V24 Prime95). I.e., I ran P-1 stage 2 on exponent Mnnnnnnnn to a B2 limit of 100000000, then when that didn't find a factor I extended the stage 2 from B2=100000000 to B2=200000000. Then I ran more stage 2 to extend B2 from 200000000 to 300000000.

Problem: V5 PrimeNet credited me with CPU time for each of those P-1 results apparently as though each of them had been run from the very beginning -- i.e., first for stage 1 time to 12345678 and stage 2 time from 12345678 to 100000000, then for stage 1 time to 12345678 and stage 2 time from 12345678 to 200000000, then for stage 1 time to 12345678 and stage 2 time from 12345678 to 300000000. This, apparently, triple-credited me for the time for Mnnnnnnnn stage 1 and for the first third of Mnnnnnnnn stage 2 time, and double-credited me for the second third of Mnnnnnnnn stage 2 time.

(The above is just an example. I actually reported about 30 stepped stage 2 results for the same Mnnnnnnnn exponent, thus getting a 30-times-what's-deserved credit for my Mnnnnnnnn stage 1 time, and so forth ...)

I'm gonna (undeservedly) rank pretty high in P-1 ... in a few minutes when the next hourly report is generated.

Edit: Bingo! The V5 report makes me look like a P-1 fiend -- up 26 in P-1 rank (up 57 overall) in 1 day, apparently with over a GHz-month thrown at each of my supposed multiple unsuccessful attempts. (Well, I am a P-1 friend, but not that much.)