Possibly unnecessary TF by LordJulius

[Andrew Usher]

I am sure you did. Ironically, there's now a stronger case for poaching the as just _one_ remaining exponent will reach the milestone now, and that's reported no progress for at least 2 weeks (and has that long left to run, also) - yet only you, the diligent user, got poached.

[Madpoo]

Originally Posted by Happy5214

That's the key part, and I should amend what I said previously. I don't know enough to know whether azhad was actively working on the exponent (and reporting that progress) when LordJulius decided to start his PRP test. If active progress was being reported, then poaching the exponent is indeed unacceptable.

I'm very late to the game on this debate, but...

I went back and looked at azhad's history of reporting results to the server for that exponent M62798611. I don't know why LordJulius decided it wasn't being worked on because we do keep a history of when users check in their progress (the interim residues that it sends in every 5M iterations by default).

It started out with the first 500,000 iteration actually, around Oct 2022, so it was assigned before that at least.

After around Dec 20, 2022, it kicked into gear and started in earnest, doing each 5M iterations about every 1-2 days. There were some fits and starts where it didn't report in (like between Dec 27 and Jan 6 and again a big gap from Jan 12 to Feb 6) but it did keep mostly steady progress after that when it reported 30M iterations on Feb 6. It had been reporting in almost daily after that with steady progress.

Anyone looking at that exponent in the assignment page would have seen the same thing and seen it had reported in recently. Unless LordJulius started his PRP test before Feb 6th and then never checked later to see if azhad had checked in again, he should have seen it.

At any rate, azhad's assignment didn't expire. We do have the assignment rules that can be a little confusing at first glance, but algorithmically it's very deterministic. In this case, even though this exponent was a cat zero at the time it was done, it was assigned long ago enough that it would have been in a different category at the time. You have to look back at what that was when assigned, not as if it had just been assigned when it was already cat 1 or zero.

I'm going to guess it was category 2 when assigned (if it was assigned just prior to Oct 18, that would be the case). Cat 2 gets 120 days to complete. 4 months on from Oct 17th (just assuming it was around then) would put the expiration pretty close to when he turned in his result on Feb 13th.

The keen observer will see that on the assignment page, I've done my best job to estimate the "time to live" (TTL) of an assignment. It tries to take into account those variables like when it was assigned, the current date, whether it has started and/or checked in recently, etc.

I do remember seeing the countdown to that <63M milestone and was surprised to see someone had done a PRP on it, since the assignment hadn't expired and I was pretty sure it was making progress. When we get close to those milestones, I tend to watch it more because when we cross the last one, I update that milestone page with the new info. I didn't pay attention at the time to who poached it, but I did think to myself that it was probably going to cause some comments here and I forgot to come check until I saw this thread.

If I could offer some constructive criticism, I'd say that if you're going to do something like this, understand that if you get it wrong and poach an assignment that someone is actively working on, you're going to upset them and ruffle a few other feathers along the way. I'm not innocent about this either - I've done my share of poaching back in the day (late 90's I reckon) and I think I'm the one who came up with the term "poaching". I tried to make sure I was getting abandoned assignments, but I got it wrong often enough that I see how it was very annoying to people, especially the ones who had the assignments.

So... learn from this and do better.

[Madpoo]

I wanted to amend a little of my research. When I was looking at the interim results, I didn't realize that the one I saw from Oct 2022 was actually a different user. azhad actually checked in the first 500,000 interim residue on Dec 20, 2022.

What's more, I had a copy of the database from before the assignment was turned in and I can see it was actually assigned on "2022-12-17 21:14:03.413"

That would have made it category 1 when assigned which gives it 60 days to complete (that would have been "2023-02-15 21:14:03.413"). Cat zero on that particular day was 62724524 or less, so it was close, but his assignment was definitely Cat 1.

He turned in a result on Feb 13th, so I can say confidently it was not expired when it was poached. I mean... when the poached result came in on Feb 12, that assignment still had 2-3 days left on it, and I know that just the day before, on Feb 11th, azhad had crossed the 50M iteration point.

[LordJulius]

Originally Posted by Madpoo

I wanted to amend a little of my research. When I was looking at the interim results, I didn't realize that the one I saw from Oct 2022 was actually a different user. azhad actually checked in the first 500,000 interim residue on Dec 20, 2022.

What's more, I had a copy of the database from before the assignment was turned in and I can see it was actually assigned on "2022-12-17 21:14:03.413"

That would have made it category 1 when assigned which gives it 60 days to complete (that would have been "2023-02-15 21:14:03.413"). Cat zero on that particular day was 62724524 or less, so it was close, but his assignment was definitely Cat 1.

He turned in a result on Feb 13th, so I can say confidently it was not expired when it was poached. I mean... when the poached result came in on Feb 12, that assignment still had 2-3 days left on it, and I know that just the day before, on Feb 11th, azhad had crossed the 50M iteration point.

Thus, on the Active Assignments page, the "Assigned Date" column is useless unless one also knows what Cat it was in when assigned?

So, there is no progression of requirements from 'Cat when assigned' to 'Cat as of current date' ?
(Except as given in the last part of Assignment Rules sections for Cat 2/3/4... which actually never come into effect because of the associated hard time limits... unless "completed in as little as ___ days' does *not* mean 'get it done within ___ days'... in which case, what does it mean?)

-----------------------------------------------------------
A little different question - example exponent:
Exponent Work Type Stage,% Expires ETA Assigned Last Update Next Update Est Complete
85296919 PRP -66 87 2022-12-20 2023-03-03 2023-03-10 2023-06-21

The "Est Complete" column value of [~3 months in the future] always takes precedence over the "Expires" value of [-66 days] ?

[Andrew Usher]

It is the expiration date that is correct. As long as the assignment is started (that is, has had any progress reported) that date should not change. And before you poach an unexpired assignment, you should take more trouble than you did to see that it is really abandoned.

[retina]

Originally Posted by Andrew Usher

And before you poach an unexpired assignment, you should ...

... realise that there are literally an infinity of other exponents that can be worked on instead.

There is no rush to complete any particular exponent. Find another one, there isn't a shortage of alternatives.

[ATH]

An exponent has the category it had at assignment, otherwise the restrictions "Must be completed in xx days" would be meaningless, if it could suddenly move up the categories which would lower the number of days available.

You can always check which category an exponent had by entering the assignment date in the link to the thresholds:
https://www.mersenne.org/thresholds/?dt=2022-12-20

for your M85296919 exponent, but for that exponent it is obviously Cat 4 since it is a DC more than 20M from the DC wavefront.

[storm5510]

Originally Posted by chalsall

...I, personally, find Andrew to be an amazingly effective source of noise. I suspect, but do not know, that this is true for others....

This will be off-topic: What I see as noise is looking at recent results on Primenet and seeing someone posting hundreds of TF results at the same time. I fail to understand why they hold on to them. I turn in my small batches shortly after completing them or try to.

[kriesel]

Very useful, ATH. How does one reliably determine whether an assignment was manual?

Originally Posted by storm5510

seeing someone posting hundreds of TF results at the same time. I fail to understand why they hold on to them.

How do you know they are holding on to them? Some of us run multiple GPUs, and hundreds of results might be a single day's production. Holding for a reasonable period is not an issue, and is fairly routine. People do get hospitalized, take travel vacations, get subjected to pandemic lockdowns, travel for work, etc.

[storm5510]

Originally Posted by kriesel

...How do you know they are holding on to them? Some of us run multiple GPUs, and hundreds of results might be a single day's production. Holding for a reasonable period is not an issue, and is fairly routine. People do get hospitalized, take travel vacations, get subjected to pandemic lockdowns, travel for work, etc.

It appears everyone has their own tolerance for how long they will leave their systems run unattended. As long as I am physically here, mine run, even at night. If I am going to be away more than about 4 hours, they get shut down. During spring severe weather season when there is a chance of lightning, they get shut down unless I am alert and nearby. As a retiree, I cannot simply reach into my pocket to replace anything now. What I have I must protect. The majority, I could never replace.

My apologies for being off-topic!

[slandrum]

Originally Posted by kriesel

Very useful, ATH. How does one reliably determine whether an assignment was manual

While the assignment is active you can easily check on the active assignmens page, hovering over the UserId gives you the computer name it's assigned to or "Manual testing".