[QUOTE=cheesehead;187696]It seems the only reason for that would be your impatience. Why not just let them finish? As long as they're making regular progress reports, they'll finish when they finish. Direct your attention elsewhere in the meantime. :-)

What's your definition of "straggler"? Any system slower than yours?

... and just what would be the justification for prodding them -- other than to satisfy your impatience?

... for what polite reason? ("I'm impatient for your slower system to finish its assignment" is not a polite reason.)[/QUOTE]

I'm not exactly fond of the repeated unbased accusations of me being an impatient douchebag.


My main concern is more along the lines of the possibility of somebody getting one of the last assignments and having it at the bottom of a 90+ day queue with lots of other tests ahead of it. Sometimes people even hoard excessively more than 90 days of work, but still send in status updates so it doesn't expire. At least as far as I know, there's no way to see what kind of progress a person is making like there was with the old system, so we're very much in the dark. Previously in this thread, somebody had noticed that starrynyte had the lowest exponent left, and had held the reservation for an extended period of time. Somebody kindly asked about the status of the exponent, because they liked to see the lowest exponent left slowly increase as a pseudo-sign of progress. Starrynye obliged, started giving some status updates, and there were no problems whatsoever. Obviously we're not going to googlestalk random users and threaten them to go faster if they don't finish in the next month. But if we see an exponent is assigned to a forum regular or well-known contributor (like maybe curtisc), and it's still not done in 3-4 months, I don't think they'd especially take offense to a status inquiry from people that are as enthusiastic about GIMPS as they are. Just because it would get done eventually doesn't mean it needs to take excessively long and we need to be clueless about it.

Also, this is the first time the preferred assignment system has been used. Obviously it has done a good job at helping clear the bulk of the sub-33M exponents, but in the past the major concern has been a few tests on the trailing end taking excessively long. I thought it would be a good idea to have more detailed information at how well the preferred system does at clearing out the "bottom of the barrel" relative to the past, in case there are situations in the future where there's a designated range that's being focused on. If they keep coming in 8 per day until they're all done, then great, the system works. But if we start seeing a rapid decline to only 1 or 2 a day and things drag on for months, then it would be nice to understand what the hold-ups are and how they can be prevented during the assignment stage so we won't have to deal with impatient people every time we're approaching a milestone.

[quote=cheesehead;187696][lots of ranting][/quote]
We've been through this before, as Kevin mentioned:
[quote=Kevin;187699]Previously in this thread, somebody [B][that was me :smile:][/B] had noticed that starrynyte had the lowest exponent left, and had held the reservation for an extended period of time. Somebody kindly asked about the status of the exponent, because they liked to see the lowest exponent left slowly increase as a pseudo-sign of progress. Starrynye obliged, started giving some status updates, and there were no problems whatsoever. Obviously we're not going to googlestalk random users and threaten them to go faster if they don't finish in the next month. But if we see an exponent is assigned to a forum regular or well-known contributor (like maybe curtisc), and it's still not done in 3-4 months, I don't think they'd especially take offense to a status inquiry from people that are as enthusiastic about GIMPS as they are. Just because it would get done eventually doesn't mean it needs to take excessively long and we need to be clueless about it.[/quote](bold comment mine)

So what if someone has a double check assigned to an old Pentium II and it is doing regular progress but still wont finish for another 150 days? Where do you draw the line and why?

[QUOTE=lfm;187730]So what if someone has a double check assigned to an old Pentium II and it is doing regular progress but still wont finish for another 150 days? Where do you draw the line and why?[/QUOTE]

Right now the rule is that exponents can be reassigned once they've been held by somebody for more of a year, but I believe George has stated he'd only enforce that rule if he had a compelling reason to. Having the "preferred exponent" system should prevent any system like that from being assigned an exponent near the trailing edge. Having the minimum equivalent P4 GHZ for work assignments should also help prevent those kinds of systems from getting work that will take a long time to complete. I think that this system works fine in general. Getting too heavy-handed is definitely a bad thing. But we really don't know how much better the system is going to be at cleaning up the very bottom, and if things go less than optimally I think it would be worth figuring out whether or not it's something that can be improved upon. Cheesehead seems to think people will be disheartened if we do anything at all (even if it's just asking familiar faces from the forum for a status update about why things are taking considerably longer than they should), but I think people being disheartened from what appears to be a lack of progress towards goals/milestones/"proving" something is a far more legitimate concern.

[QUOTE=Kevin;187732]... (even if it's just asking familiar faces from the forum for a status update about why things are taking considerably longer than they [b]sh[/b]ould) ...[/QUOTE]Erm, do you mean "... longer than they [b]c[/b]ould"? There is no compulsion on anyone to do any work at any time. So there [i]should[/i] be no "should".

[QUOTE=retina;187733]Erm, do you mean "... longer than they [b]c[/b]ould"? There is no compulsion on anyone to do any work at any time. So there [i]should[/i] be no "should".[/QUOTE]

I'm pretty sure I meant what I said. It seems like the big idea behind having minimum P4 GHZ requirements and reliability rankings and preferred exponents is so that exponents we want finished in a timely fashion get done in a timely fashion. You can view the preferred exponents as a kind of privilege you earn by showing you can finish work in a timely manner. If you take significantly longer than everybody else (as one might say, longer than you [i]could[/i]), you're not really holding up your end of the proverbial bargain. You received said assignment with the complicit understanding that you'd finish it in a timely fashion, so you also really are taking long than you [i]should[/i].

However, we can't justifiably take action because those understandings are just complicit, and sort of the theory behind how the system is supposed to work. It's not like there's people behind these computers twirling their mustache mischievously as they hoard preferred exponents and manually place them at the end of their 1000+ day work queue. But I could see somebody reserving a large number of assignments on "trusted" machine and spreading them around to their slower machines with no real malicious intent. If there are still enough people taking a long time to finish preferred exponents (and there very well might not be), then it would be worth considering what additional implementations could be used to ensure steady progress on the trailing edge. Maybe something along the lines of adding a work-type where the "agreement" is made more explicit, and users are informed the assignments can be reassigned after less than one year.

But again, this is all predicated on the idea that there is some problem still remaining with the current system with respect to the very end of the trailing edge, and even whether or not there is a problem will be a debatable point. We know the system does a good job at taking care of the bulk, I'm just not entirely convinced that it will completely remove all the exceptional cases that hold things back. At this point, I'm just suggesting that if people still have an interest in tidying up the trailing edge, then we're near the point where it would be a good idea to pay extra special attention so we can identify possible issues/loopholes and proactively address them instead of dealing with the same crap every time we get close to a milestone.

Kevin
 

I like this guy. He would be great on a debate team. ;-)

[QUOTE=Kevin;187742]If there are still enough people taking a long time to finish preferred exponents (and there very well might not be), then it would be worth considering what additional implementations could be used to ensure steady progress on the trailing edge.[/QUOTE]

You seem to think there is some sort of "rate of progress" that would be agreeable to at least a majority of those concerned. The trouble is nailing down what that rate actually is would be contentious in itself. For myself I'd tend to say the current 1 year limit is just fine thank you and even if the "mark" you watch only moves once a year THAT IS STEADY PROGRESS still.

Note: we are over-subscribed with Trial Factoring and under-subscribed with Double Checks. Logically this means we should move some more slower machines from TFing to DCing. I think this would probably make the DC stragglers even more visible.

[QUOTE=lfm;187836]You seem to think there is some sort of "rate of progress" that would be agreeable to at least a majority of those concerned. The trouble is nailing down what that rate actually is would be contentious in itself. For myself I'd tend to say the current 1 year limit is just fine thank you and even if the "mark" you watch only moves once a year THAT IS STEADY PROGRESS still.[/QUOTE]

No, I very clearly stated "whether or not there is a problem will be a debatable point." And progress isn't guaranteed after a year. The exponent could theoretically be assigned to somebody else that takes an additional year to finish.

Also, imagine people who might make it their goal to clear out the lower end. It's a little disheartening if you spend a lot of time working on clearing everything beneath a certain range, and it takes an extra year to actually reach the goal because a few people take their time. There's never really a sense of accomplishment or moment of triumph for your wok when you get within arm's length of the goal, but things don't officially end for another few months.

[quote=Kevin;187699]My main concern is more along the lines of the possibility of somebody getting one of the last assignments and having it at the bottom of a 90+ day queue with lots of other tests ahead of it.[/quote]So?

[quote]Sometimes people even hoard excessively more than 90 days of work, but still send in status updates so it doesn't expire.[/quote]So what is it to GIMPS?

[quote]Obviously we're not going to googlestalk random users and threaten them to go faster if they don't finish in the next month.[/quote]Then why do you need more information on their progress? So someone can poach their assignments?

[quote]Just because it would get done eventually doesn't mean it needs to take excessively long and we need to be clueless about it.[/quote]"excessively long" by what standard? All I see you using is your own impatience as that standard. GIMPS suffers no harm by having someone take, say, a year to complete an assignment. I see no mention by you of any standard of excessivity other than your own sense of impatience, so I'll keep calling it like it is.

[quote]how they can be prevented during the assignment stage so we won't have to deal with impatient people every time we're approaching a milestone.[/quote]So, you value the impatience of a vocal few over the quiet sustained contributions of those with systems deemed "too slow" by your standards?

Please describe exactly what REAL damage would be done to GIMPS by someone who, though genuinely making steady progress, does not meet your standards of patience. What is it, exactly?