How to reserve very low TF work

[TheMawn]

Originally Posted by LaurV

They did hit every single factor below ~54 bits (this we can confirm), doing "by k" search

You can't know this. It's very likely that they did, but how do you know they didn't skip and or miss any k's?

[TheMawn]

Originally Posted by snme2pm1

Apparently TJAOI is doing more than one kind of thing.
A contrasting example voiding your statement is: http://www.mersenne.org/report_expon...=899893&full=1

Well if they're doing some P searching then yes a small amount of TJAOI's work is being recorded as explicit no factors from X to Y. But all the by K searching that got us talking about TJAOI in the first place finds factors only, no "No factors" results.

[snme2pm1]

I don't know why or how, but Prime95 did recently chose to register an assignment for an otherwise unassigned job in flight on a local machine.
I have heard past mention of this as a possible course, but have never witnessed it before.
I have been puzzling as to what circumstances might have coaxed it along that path, without success.
Despite various efforts, choosing apparently lonely exponents, I have not been able to see it happen again, which is confusing.
Any ideas how to make it happen?
From the log:

[Thu Feb 12 22:10:38 2015 - ver 28.5]
Registering assignment: Trial factor M850637
Assignment registered as: 71B5E36C503FCCA1996ABEA6A2FDF37F

[Prime95]

Originally Posted by snme2pm1

I don't know why or how, but Prime95 did recently chose to register an assignment for an otherwise unassigned job in flight on a local machine.
I have heard past mention of this as a possible course, but have never witnessed it before.
I have been puzzling as to what circumstances might have coaxed it along that path, without success.
Despite various efforts, choosing apparently lonely exponents, I have not been able to see it happen again, which is confusing.
Any ideas how to make it happen?

If there are any active (or expired?) ECM assignments then the server won't register the TF assignment. I also don't know if ECM assignments ever expire.

[snme2pm1]

Originally Posted by Prime95

If there are any active (or expired?) ECM assignments then the server won't register the TF assignment. I also don't know if ECM assignments ever expire.

Since the notion of "expired" has been emphasised, gazing at
http://www.mersenne.org/report_expon...exp_hi=&full=1
It appears that it has the most rare distinction of not having any expired assignments listed.
For contrast, a look at
http://www.mersenne.org/report_expon...o=77557&full=1
shows five expired assignments, all expired more than three months ago.

The description at http://www.mersennewiki.org/index.ph...d_test_results, would suggest that exponents with expired assignments are candidates for in-flight registration.

Is this cleverly designed Web API registration mechanism now entirely thwarted due the those expiry records being retained for a long time, perhaps indefinitely?

[Prime95]

Originally Posted by snme2pm1

Is this cleverly designed Web API registration mechanism now entirely thwarted due the those expiry records being retained for a long time, perhaps indefinitely?

I looked at the register assignment code on the server. It looks like an expired assignment will not affect registration. An active ECM assignment will.

If this happens again, post the error code and text. It might help narrow down the cause.

[snme2pm1]

Now I know much more than when I posed the question.

I once would have liked to trust http://www.mersennewiki.org/index.php/Worktodo.txt, but it is misleading by not indicating that the assignment ID portion is optional.
This has led to compounding of confusion.
Some time ago I learned that Prime95 tolerates “N/A” as the AID, but apparently that is also incomplete and misguided thinking.
Far more potent is leaving the AID field void and eliminating the field delimiter comma.
After reading the sources of Prime95, particularly commonc.c, it is now clearer what really happens.
The “N/A,” thing is an optional prefix to an optional AID field; they can both be present.

Prime95 won’t attempt to register a work unit where “N/A,” is present.
So for Prime95 registration purposes, any “N/A,” must be removed.
I suspect that at some time Prime95 chose to avoid repeated attempts to acquire registration after a failed attempt; preserving that failed state by the presence of “N/A,”.

But more, why does the PrimeNet agent, http://www.mersenne.org/report_factoring_effort, now purposed for building direct worktodo txt content, inject a “N/A,” at all?
A short test (without reading source) suggests that both mfaktc and mfakto work quite compfortably without that noise.
So for whom or what purpose was that injected?
Is there any other agent where it is meaningful?
It seems to me that “N/A,” was and remains state information strictly for the private purposes of Prime95.
One odd user exception might be for queueing a Prime95 work unit with registration deliberately thwarted.