Newer milestone thread

[chalsall]

Originally Posted by Prime95

The register codepath may not be adhering to the assignment rules.

Perhaps it should.

[Madpoo]

Originally Posted by Prime95

The get-me-a-new-assignment and register-this-assignment-from-worktodo has two different code paths on the server. The register codepath may not be adhering to the assignment rules.

It's probably allowing the assignments to be made because:
a) no existing assignment
b) the machine already has a save file (even if just a single iteration)

In such a case, when the computer contacts the server, it will always generate a new assignment for that exponent, regardless of categories, previous expirations, whatever.

I could be wrong, but that seems like the most likely explanation.

When I have time I can check a few things... I was on vacation and just got back the other day so I haven't had a chance yet.

[kriesel]

https://www.mersenne.org/assignments...00000&exdchk=1

200 to go at the moment, so presumably Cat 0 rules would apply; only 34 (17%) are PRP
46 (23%) patgie all LL; computer n/a; with peculiar expiration and reassign pattern, progress but none indicated on exponent status along the way
21 WR (all PRP) no updates since 2019-07-22
18 Daniel Procopio all LL; computer n/a; (and 2 of these now show 80%+ progress) last updated 2019-08-09
7 -Anonymous- all LL (only 1 with a current update, most weeks old) mostly computer: n/a
6 arnaud all LL (4 have not reported progress in weeks)
5 curtisc all LL (all updated recently)
4 Jon Pace (all LL) frequent and current updates
3 kriesel all PRP on prime95, daily progress updates
3 Stephan Grupp 1 last update 2019-08-20, 2 2019-08-14
subtotal 113 among top 9 usernames as holders of milestone assignments with > 1% each
Nearly half, 95 of the remaining exponents to the 86M milestone are both not reporting progress in weeks, consistent with being stalled, and are held by the 5 usernames with the most remaining exponents, out of about 90 usernames listed.

[rudy235]

Originally Posted by kriesel

https://www.mersenne.org/assignments...00000&exdchk=1

200 to go at the moment, so presumably Cat 0 rules would apply; only 34 (17%) are PRP
46 (23%) patgie all LL; computer n/a; with peculiar expiration and reassign pattern, progress but none indicated on exponent status along the way
21 WR (all PRP) no updates since 2019-07-22
18 Daniel Procopio all LL; computer n/a; (and 2 of these now show 80%+ progress) last updated 2019-08-09
7 -Anonymous- all LL (only 1 with a current update, most weeks old) mostly computer: n/a
6 arnaud all LL (4 have not reported progress in weeks)
5 curtisc all LL (all updated recently)
4 Jon Pace (all LL) frequent and current updates
3 kriesel all PRP on prime95, daily progress updates
3 Stephan Grupp 1 last update 2019-08-20, 2 2019-08-14
subtotal 113 among top 9 usernames as holders of milestone assignments with > 1% each
Nearly half, 95 of the remaining exponents to the 86M milestone are both not reporting progress in weeks, consistent with being stalled, and are held by the 5 usernames with the most remaining exponents, out of about 90 usernames listed.

Suggestions? Solutions?

[Prime95]

Originally Posted by rudy235

Suggestions? Solutions?

Patience. The world won't end if this "milestone" is delayed a month, two months, or a year.

[chalsall]

Originally Posted by Prime95

The world won't end if this "milestone" is delayed a month, two months, or a year.

No, but... The whole point of the assignment rules was to lessen the motivation for poaching, which /does/ slow down GIMPS' progress...

[Prime95]

Originally Posted by chalsall

No, but... The whole point of the assignment rules was to lessen the motivation for poaching, which /does/ slow down GIMPS' progress...

Agreed.

I personally don't understand the obsession with achieving a particular milestone as quickly as possible. That said, GIMPS motto has always been to do what you consider fun.

[rudy235]

Originally Posted by Prime95

Patience. The world won't end if this "milestone" is delayed a month, two months, or a year.

I did NOT mean emergency solutions.. Just constructive ideas of some sort.

[Prime95]

Originally Posted by Prime95

The get-me-a-new-assignment and register-this-assignment-from-worktodo has two different code paths on the server. The register codepath may not be adhering to the assignment rules.

I cobbled together a possible solution. Hopefully, I've not broken anything.

[LaurV]

Originally Posted by Prime95

The get-me-a-new-assignment and register-this-assignment-from-worktodo has two different code paths on the server. The register codepath may not be adhering to the assignment rules.

Interesting. Any rationale for it? (I am thinking along the line "let's be polite with the user, if he wants to work on this exponent and only on this, we can't stop him anyhow" hihi, but as we are not in a perfect world... sheep happens sometimes, like in the past there were cases where the user installed P95 in ProgramFiles and he then didn't have rights to modify worktodo, for example, requesting the same assignments again and again, so the real questions are how do we avoid this, and how much effort should be put in avoiding it, I personally don't make a big deal from milestones, and I think there are exponents enough for everybody to pick whatever he wants, as far as any progress happens. Someone I know here said that 101 miles per hour is better than 100 miles per hour...)

[kriesel]

Originally Posted by Prime95

I cobbled together a possible solution. Hopefully, I've not broken anything.

Not to pick on patgie, who may have actually done the project a service by sort of exposing an issue, but a rough extrapolation based on 12 exponents (21%) finished or lost to others in 11 days as an exponential decay gave about 6 more months to reaching a single remaining assigned exponent, in addition to the 8 months already past since initial assignment.