Lost assignment?

[mdettweiler]

Originally Posted by ATH

As was discussed in this thread: http://www.mersenneforum.org/showthread.php?t=13472 all LL tests have random shifted S0 start value, so the test will count as a doublecheck even if its left at Test= in worktodo.txt

Yes, but what I was thinking is that by changing it to Doublecheck=, Primenet could register the test as a doublecheck to "officially" be run alongside the first-pass test. That way, the OP can set his client to send end dates at a reasonable interval without the client talking to Primenet and thinking "hey, this isn't assigned to me, I'll delete it now".

[NBtarheel_33]

Originally Posted by mdettweiler

Yes, but what I was thinking is that by changing it to Doublecheck=, Primenet could register the test as a doublecheck to "officially" be run alongside the first-pass test. That way, the OP can set his client to send end dates at a reasonable interval without the client talking to Primenet and thinking "hey, this isn't assigned to me, I'll delete it now".

I tried this just yesterday actually, with one of the remaining couple of numbers below 30402451. It doesn't work, that is, even though I used a DoubleCheck= line, the assignment key still came up as N/A.

I might add that I technically wasn't trying to maliciously poach, as the assignment of the number in question is something like 17 months old, beyond the 1-year assignment expiration threshold.

[lfm]

I still think dominicanpapi82 should give up on this exponent. He lost it when he didn't set up his machine and monitor it properly. He should start a new one.

The server will not support two outstanding assignments on the same exponent even if they might not overlap such as a LL and a DC. Normally there is plenty of work for everyone and there is no need to double up.

[mdettweiler]

Originally Posted by lfm

I still think dominicanpapi82 should give up on this exponent. He lost it when he didn't set up his machine and monitor it properly. He should start a new one.

The server will not support two outstanding assignments on the same exponent even if they might not overlap such as a LL and a DC. Normally there is plenty of work for everyone and there is no need to double up.

Normally I would agree, but in this case the 25% already completed represents one year of CPU time. A CPU-year is a whole heck of a lot of time--consider, for instance, just how many 50M exponents could have been crunched in that amount of time. We wouldn't throw away the equivalent number of completed 50M exponents, so why do so with this?

[NBtarheel_33]

Originally Posted by mdettweiler

Normally I would agree, but in this case the 25% already completed represents one year of CPU time. A CPU-year is a whole heck of a lot of time--consider, for instance, just how many 50M exponents could have been crunched in that amount of time. We wouldn't throw away the equivalent number of completed 50M exponents, so why do so with this?

Took the text right off my keyboard, LOL. This is exactly how I feel about it. Given that the exponent wasn't reassigned until April 25, which is only two months ago, and he is 25% of the way done - by xorbe's experience, dominicanpapi is *at the very least* 2-3 months ahead of this new assignee. If you're troubled by the ethics of possibility of stepping on the new assignee's toes, dominicanpapi could simply finish the result and wait to submit it as a double-check. By and large, chances are it isn't prime, anyway. Chances also are high that the re-assignee (I notice it's an S-series ID, which is one automatically generated by PrimeNet - indicating that the assignee may have even signed up for GIMPS accidentally as part of a stress test!) might abandon the test long before its completion (remember that many newbies are going for the gusto on the EFF prize, unaware of just what is involved in testing a number of that size!).

If we were talking about 25% of a run-of-the-mill 50M exponent, I'd maybe say go ahead and toss it (and even then, it seems silly not to finish it as a double-check). I've lost large chunks of 40M and 50M assignments on borged machines, and it's quite annoying. But given the time involved in, as well as the chance of the reassignee never finishing, a 332M assignment, I'd say throwing away a quarter of that work is definitely doing GIMPS more harm than good.

Hells bells, *I'll* gladly take your save file, and let it finish at a slow enough pace to be a double-check (or at least give the re-assignee his year to finish, which by the way, is now only 10 months).

[henryzz]

IMO the likelyhood that a third test will be required anyway is quite high considering how long these tests are taking. The shifted doublecheck will probably be done with a much faster machine much later that will be much more likely to be correct. Hopefully then one of the currently running tests will match. I personally doubt both will. Can anyone say how likely it is that this test will be errorfree?

[JuanTutors]

Originally Posted by henryzz

IMO the likelyhood that a third test will be required anyway is quite high considering how long these tests are taking. The shifted doublecheck will probably be done with a much faster machine much later that will be much more likely to be correct. Hopefully then one of the currently running tests will match. I personally doubt both will. Can anyone say how likely it is that this test will be errorfree?

All the double checks on my previously done exponents on this computerhave matched, so probably higher than a random sample. I don't know the complete answer to your question, though.

//EDIT: And I agree with you that, given the number of 10M LL tests and double checks that don't match, there is definite value in not discontinuing a test that will complete circa 2013/2014.

Given the amount of time it takes to complete a first time test, which is the same order of magnitude as the life of a computer and one order of magnitude below the average HUMAN lifetime, I would even go as far as saying that we should start two tests of an exponent at the same time and modify the prize money allotment accordingly. I hate to take this decidedly morbid and extremely tangential route, but heck, the probability of DYING between when you start a test and when it's double checked, assuming it's double checked checked immediately, is hovering around 3% on average!

[cheesehead]

Originally Posted by dominicanpapi82

I hate to take this decidedly morbid and extremely tangential route, but heck, the probability of DYING between when you start a test and when it's double checked, assuming it's double checked checked immediately, is hovering around 3% on average!

Let's simply turn this into a reminder to all of us to ask ourselves how long it's been since we last updated our last will and testament.

[JuanTutors]

Originally Posted by cheesehead

Let's simply turn this into a reminder to all of us to ask ourselves how long it's been since we last updated our last will and testament.

Or whether we have one! Hah!

[cheesehead]

There's no better time than now to start.

Start by listing what will be required in your will.

The following sites are about how to write your own will, but what I'm saying is just read about the various data items (such as, but not limited to, how you want your stuff divvied up), and start listing them.

http://www.mahalo.com/how-to-write-a-will

http://www.wikihow.com/Write-Your-Ow...-and-Testament

Don't actually write the will; just write a list of what will go into the will. You'll need that list when you go to a lawyer to have the will written for you, anyway.