Re: Mersenne: Poaching -- Discouragement thereof

[Mary K. Conner]

I also have a proposal. This would need to be for the overhaul of Primenet.

1. When someone who has not held an assignment in the last year checks it in completed (in other words almost always a poacher), that submission is not acknowledged publically in any way. The exponent remains listed on the AER, no entry is made to the completed exponents page, the poacher receives no credit, and if the exponent expires before the original assignee finishes it (or is checked in with no progress made), the original assignee receives the public credit for it, and only the private database notes the poacher. There are a number of poachers who could care less about credit, but they do want to see those entries fall off the AER and milestones reached. If that didn't happen, and if it were impossible to even tell if another poacher had already poached an exponent, I believe these poachers would give up. It might not stop those who dash off an exponent just before it expires, but only if they don't care about credit.

2. The vast majority of "exponents out for a long time" happen because people who only have their computers on a very short time every day have exponents hang out in their queue for a very long time because Prime95 has a hard time dealing with estimating true times to completion when computers are on for short times or not turned on very often. If someone takes a year to complete an exponent, that's no big deal, but if they have five exponents in their queue (because Prime95 doesn't realize it will take five years to finish them all), it will be five years to finish the last. The best way to deal with that (even with the existing client base) is for Primenet to estimate the client's true progress and to take back those exponents that report no work done (IOW those that are in the queue behind the exponent that is being worked on) when they do check in (giving them an "exponent not assigned to this computer" error), and then give that client exponents on the leading edge when it asks for some to replace those errored out. That way when the year is up and that first exponent is completed, the next exponent that computer starts on will be one is much closer to the leading edge.

3. I like that Primenet should choose whether to give an expired exponent to a computer based on its history. However, I would like to suggest that the expiries that are the very closest to the trailing edge (perhaps the bottom 10%, if there are 4000 of that type outstanding, any expiries in the bottom 400 would fall in this criteria) should go to computers that not only pass Primenet requirements but whose owners have explicitly decided to ask for them. It should be an advanced checkoff or undoc option. It should be suggested to only use this option for computers that the GIMPSter has regular direct access to.

4. Primenet requirements for getting expiries in the bottom half of the active range should include using version XX.XX of Prime95, which has as a feature that it automatically sorts the worktodo list when rewriting it, placing the lowest exponents to be completed first. Exponents that have work in progress should sort to the top, and the sort should only be within work types. If one is working on factoring, and decides to switch to doublechecks, the doublechecks should not be popped above the factoring assignments or the factoring assignments will be a long time getting done (at least until Primenet perceives that a situation like 2 above is happening and reclaims the factoring assignments in the queue).

5. Version XX.XX of Prime95 and the overhaul of Primenet should support factoring limits and checkins of partial factoring progress. Right now if an exponent starts at 58 and needs to go to 66 bits, and someone factors through 65 and then expires that exponent, the new assignee has to start all over again at 58. The undoc factoring limits option should work with Primenet, and Primenet should preferentially give work over 64 bits to SSE2 clients, and perhaps direct others to factor only up to 64 bits unless there aren't enough SSE2 clients to handle the over 64 bit work (or if the owner of a machine asks for over 64 bit work).

Okay, I'll stop. I keep thinking up things. :)

_________________________________________________________________________
Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers