thoughts about assigmments ques

[Keroberts1]

would it be possible t ohave the server check how many residues are already available for each tests before it is assigned? I noticed that the gartbage account has alot of tests in it that have already had residues returned once. I think this happens because someone who had the test before but never finished it started it up again and returned a result after the test had already been reassigned and finished. Also at the same time the admins could assign a residue value to all of the remaining N values of primes so that once a test is abandoned by a user instead of going to the dropped test que it'll simply be dropped for good.

[vjs]

I think the simple solution to what your saying is basically do a sanity check on the test before it is added to the dropped test que.

The first thing to do would be create a dat list which only contains k/n pairs which require more testing.

The dat list would continually decrease in size by two criteria:
1. If a factor is found remove that k/n pair from the (dat/server que).

I know this happens, since when Joe and I submitted a large number for factors at one time. The number of tests in the second pass que reduced by the number of tests complete that day plus the number of factors we submitted. This was one of the ways we knew our secondpass sieve effort was actually working and those factors were actually previously missed.

2. If two matching residues are found remove that k/n pair from the (dat/server que).

I don't think this is happening, but it would slove almost all que issues if the following were done.

When a test is dropped perform a sanitiy check before are added to the dropped test que.

Does the k/n match one of the k/n in the dat list? No discard

The only issue would be an early double check, and a tripple check if the at fault/dropping computer ever finished the test. Also if someone were to invent a test it would never enter into the system (A common problem).

[DigitalConcepts]

Originally posted by vjs
2. If two matching residues are found remove that k/n pair from the (dat/server que).

I'd also would want to make sure the residuals came from different people - but that might be a later process.

[vjs]

Well considering that the only residues which are considered in the process are those which are server assigned*.

*- Currently if you create your own test, even if its a 100% honest effort. (actual, k/n pair, tested 100% correct, yada yada). The server will store the residue and check if the result is prime. But won't consider it for either scoring or as a completed test removing that pair.

The above helps avoid problems and that the test is actually tested. So the need for actually checking that residues are from different uses at this point isn't really needed.

It's a good point to consider for the future. However with the size of these tests and the user base. The chances are pretty small the same user actually gets assigned the same test twice.

[Theadalus]

Originally posted by vjs
The chances are pretty small the same user actually gets assigned the same test twice.

I did!

[The first time it was assigned Aug 18 and the second time Sep 17, so there wasn't even a month between it!]

[Matt]

Those people with a higher chance of getting the same test twice are probably those running large herds, in which case the test is probably run on a different machine and has the same affect as being from a different user. The key thing really is the same residue from two different runs of the test on two different machines.

[Keroberts1]

well the meaning of not giving it to the same user it to prevent a user fro msumbitting a fake residue twice or having a saved version of the tests and simply finishing from the saved test which already hasan erro and could reproduce the same error. It isn't likely to happen but people do try to cheat stats sometimes.

[DigitalConcepts]

Originally posted by Keroberts1
well the meaning of not giving it to the same user it to prevent a user fro msumbitting a fake residue twice or having a saved version of the tests and simply finishing from the saved test which already hasan erro and could reproduce the same error. It isn't likely to happen but people do try to cheat stats sometimes.

Thank you for saying in a clearer way, what I meant. If there is a way to get ahead quick - you can count on someone to do it.

[Theadalus]

I expected not to get any credits/points for the second time i submitted the same test, but i did (in this case it would be fair because i got it assigned twice).

When you keep track of how many times and to whom a test is assigned, you can filter out cheats. When a person submits a test twice while he/she got it assigned once, the second time no credits/points!