Primenet web design

[Madpoo]

Originally Posted by chalsall

Thanks for that reminder.

One of the first things I have my people read is Brooks' The_Mythical_Man-Month. It just makes comms amongst geeks faster.

Please ask if there are any questions.

There's also the sunk cost fallacy to consider... just because Primenet currently handles Fermat work doesn't mean we should spend more time on that aspect of it.

[James Heinrich]

I have zero intention of adding any tracking of effort (ECM, P-1, etc) but since I had the factor data at hand I have put up a simple Fermat Factors summary page on my site:
https://www.mersenne.ca/fermat.php
If you have any issues with that page, please raise them in the mersenne.ca forum rather than this thread.

[cxc]

Hi James (and other mersenne.org web gurus), apologies for re-iterating my bug report. The fix for exponent 2 (incorrectly being tagged as Fermat F1) doesn’t appear to have taken; I assumed before that I was looking at a cached copy which wouldn’t update because of browsers at my end being finicky, but having cleared caches and reloaded I’m still seeing the attached.

[James Heinrich]

Originally Posted by cxc

apologies for re-iterating my bug report

Don't apologize if something not working correctly

I'm not sure what happened, clearly my change got clobbered somewhere, but I have re-applied that fix.

[Andrew Usher]

At this point, as no defence as even been offered for P+1 but not P-1 being accepted, it must be considered a bug, and a non-trivial one. The only possibility excluding it from that category would be that it's being intentionally kept to spite me, and I wouldn't believe that of James, Aaron, George, or presumably anyone involved with this.

[James Heinrich]

To date, nobody has submitted (to me anyways) any P-1 result line that they have tried to submit on the manual results page but wasn't accepted.

Anything that is Fermat related is somewhat low-priority because, well, GIMPS, but I'll take a look if someone has a specific example.

[Dr Sardonicus]

Originally Posted by Andrew Usher

At this point, as no defence as even been offered for P+1 but not P-1 being accepted, it must be considered a bug, and a non-trivial one. The only possibility excluding it from that category would be that it's being intentionally kept to spite me, and I wouldn't believe that of James, Aaron, George, or presumably anyone involved with this.

I don't know anything about Primenet web design except what I read on the Forum. I'm not a programmer.

But I note the following WRT the above quotation:

1) The spelling "defence" is British, not American. The American spelling is "defense."

2) The use of the word "defence" connotes an accusation of deliberate wrongdoing. This idea is echoed later on [see (4)].

3) The first sentence uses the passive voice three times ("been offered," "being accepted," "must be considered.") The second sentence adds a fourth use ("being kept.) This is aggravated stuffy prose.

4) The second sentence has a false dichotomy - it's either a bug, or it's something the programmers are doing intentionally to spite me personally. Why some random user would think themselves important enough to be worth the effort of writing the code simply to spite them is quite beyond me. From my perusal of the Forum, it appears that the problem (P+1 accepted, P-1 rejected) has been noticed by other users. If this is correct, it effectively rebuts any suggestion of spite against any individual user personally.

Not being a programmer, and not being familiar with how users make requests, I don't know that the problem isn't with the input. AFAIK the problem could be a parameter being set wrong or something. I also don't know the system requirements for the two types of request. For all I know, a user's system could be acceptable for a P+1 but not a P-1.

Absent any evidence of malicious programming, the aggrieved user might better request an explanation rather than a "defence," and otherwise to be polite and show respect to the people who actually write the code. If you give them a summary of your system, the request you sent, and how the program responded, IMO that would go a lot farther toward resolving the problem, than the approach the above quote takes.

[kriesel]

Originally Posted by Dr Sardonicus

If you give them a summary of your system, the request you sent, and how the program responded, IMO that would go a lot farther toward resolving the problem, than the approach the above quote takes.

Exact details of inputs and outputs; copy/paste, screen captures, typically, are productive, while unspecific kvetching is not. See #22 here.

Why some ... user would think themselves important enough to be worth the effort of writing the code simply to spite them is quite beyond me.

(With the implication, "important enough, to the code authors, in a negative way") That it's an inexplicable conclusion is a good thing, considering what might be behind generating such beliefs.

For all I know, a user's system could be acceptable for a P+1 but not a P-1.

AFAIK the system requirements of P+1 and P-1 to similar bounds are comparable for similar exponents. Frequency of use is quite different though. For Mersenne numbers, P-1 is very common, almost universal, while P+1 is rare, since P-1 is an efficient means of avoiding full primality tests sometimes by finding a factor, while P+1 is less efficient and so not included in the normal progression of production factoring and testing at the wavefront, and half the time P+1 worktodo is actually effectively P-1 as performed. There's more in the prime95 v30.4/5/6 forum thread and prime95's v30.6 section of whatsnew.txt.

[chalsall]

Originally Posted by Madpoo

There's also the sunk cost fallacy to consider... just because Primenet currently handles Fermat work doesn't mean we should spend more time on that aspect of it.

I once told multiple bosses (after they had spent tens of millions of dollars) that the submarine fibre optic cables they paid for were, by definition, a sunk cost.

For some weird reason, they didn't find it at all funny. I giggled.

[Andrew Usher]

It is worth emphasising what Kriesel just alluded to, even though we should all know already, just how big a disparity there is between P-1 and P+1 in favor of P-1, and the same holds for Fermats and any other number of cyclotomic form (but for Lucas sequences where there are as many P+1 as P-1 factors, it will be different). Thus rightly George gives a low priority to adding the polynomial stage 2 to P+1 (I am sure that they are different enough there that the code needed would differ substantially).

In any case, I did not actually expect a reply to my last, unless it were to state that the problem has actually been fixed, and no substantive arguments have changed so I will not be addressing any.

On another point of interest, though, the 'sunk cost' reasoning can argue in favor of, and not just against, completing a project. If it has merit, the fact that some of its cost has already been 'sunk' means that it only needs to be worth the remaining part to be justified.

[James Heinrich]

Originally Posted by Andrew Usher

unless it were to state that the problem has actually been fixed

No evidence has been provided that there is any problem.
If there is a result that should be accepted (but isn't) then provide it. If not, there is nothing to discuss.

edit:
Semantic clarification: P-1 results "should be" accepted on the manual results page if:
* they are for one of the small number of Fermat numbers tracked by mersenne.org, namely F12 through F29
* they are in standard results format from any of the usual program that produce P-1 results (Prime95, gpuowl, Mlucas, etc)