poaching aliquot sequences... - Page 2

LaurV · 2011-12-28, 03:51

I agree (and vote for) a note on the DB page. We have nothing to lose from it. It happened to me many times that DB elves or someone else factored "easy" terms in my sequences, and few times even "hard" cofactors. I got the habit to avoid reporting the factors when they were "easy" to factor, they have less then 105 digits, the sequence got a downguide or downdriver, etc. The elves will not tease a big C125-C145 that fails the ecm, so they will spent their time more useful on something else. If you report small cofactors (under C105), then give it a time of few days before you start working on it. Do something else in this time. That is why I reserved 15 sequences. I won't work all of them in the same time. For example the elves factored few cofactors under C105 on my sequence 181428, or 865752. Be my guest! Very well done! :D

This way, I don't duplicate their efforts and they don't duplicate mine (well, there are exceptions). I check the sequences with small cofactors from time to time and give them "pushes" when they seem stuck (usually ecm fail to crack it, and the elves lost their interest, that is most of the cofactors that need "a push" have some ECM already done on them - check factordb "scan" button - so I can directly schedule them for siqs/nfs after few ecm curves with high bounds).

Of course, I get angry when the elves are faster then me. That is "human". I know for example that bchaffin elves seems to have preferences for sequences with downdrivers. :D... Or yoyo's elves will not check for reservations, but only for the size of the composite. You can be sure that any C100 or lower that you report could be "lucky" sooner or later to be grabbed by such an elf. If you report it, then give it few hours/days, and work something harder in that time (which was alredy "discarded" by the elves after a couple of ecm curves, or it have too many digits to ever be considered by them).

yoyo · 2011-12-28, 09:59

The Boinc YAFU project just fetches composites from the factordb which are (currently) in the range from C80-C106 and send them as Boinc workunits to the Boinc elves.

I suggested already in some other thread, that we should introduce a possibilit to reserve a composite for 2 days. After 2 days the resrvation is removed automatic. Than we can adapt our scripts and tools to fetch only unreserved composites and reserve them.

Maybe I should chat with syd about it.

yoyo

Andi47 · 2011-12-28, 13:42

Quote:

Originally Posted by yoyo

The Boinc YAFU project just fetches composites from the factordb which are (currently) in the range from C80-C106 and send them as Boinc workunits to the Boinc elves.

I suggested already in some other thread, that we should introduce a possibilit to reserve a composite for 2 days. After 2 days the resrvation is removed automatic. Than we can adapt our scripts and tools to fetch only unreserved composites and reserve them.

Maybe I should chat with syd about it.

yoyo

Suggestion: Maybe it is even possible to extend this reservations to aliquot sequences: bchaffin's elves frequently scan the aliquot sequence threads for reservations to avoid duplication of work. Maybe the DB can do that too - or maybe bchaffin can chat with syd about that and share his scan results with the DB. Then every new composite from a reserved aliquot sequence should be automatically tagged as "reserved for two days", and any reserved aliquot sequence could be displayed with the remark "reserved, see mersenneforum (with link to the aliquot subforum)".

schickel · 2011-12-29, 00:42

Well, no poaching for a while:

Quote:

Originally Posted by FactorDB

Down for ~6 hours (replacing failed raid disk, compressing/optimizing tables, ...)

I would like to thank Syd, again, for making this increasingly valuable resource available to the entire factoring community.

schickel · 2011-12-29, 03:15

Quote:

Originally Posted by FactorDB

Down for ~6 hours (replacing failed raid disk, compressing/optimizing tables, ...)

That was quick....I must have caught that near the end of the lockout.

10metreh · 2011-12-29, 18:57

Quote:

Originally Posted by 10metreh

From looking at a few examples, poaching only seems to occur on sequences that have the downdriver (e.g. 4788 where someone poached some c13x composites) and sequences that are stable, such as the ones mentioned in this thread.
The best idea is probably never to upload your current status to the DB on these sequences, but upload up to a large composite that you have already done, so that poachers have to do more work than you do. For example, on my sequence 570280, the current status shows a c120 on line 869. In fact, I have already factored this and there is now another c120 on line 871.

Maybe I shouldn't have said that - someone GNFSed both C120s and has now advanced the sequence further than I have. I'll keep it reserved just so that I can get the credit if it terminates

Add my name to the no-anonymous-submissions petition list.

Batalov · 2011-12-29, 19:29

Just think it through well if you will be installing locks. "If there's a gun on the wall in the first act, by the third act it will be used." Example: the perpetrator will go ahead of you for a few iterations and lock a future composite for herself (yes, I am being PC

).

Instead, get and run another sequence and watch the "poached" one, -- and get the credit if it terminates. What's wrong with that?

schickel · 2011-12-29, 20:14

Well, I hope they didn't tackle the c152 on 4788. I would hate for bsquared to have wasted his work

bsquared · 2011-12-29, 20:26

Quote:

Originally Posted by schickel

Well, I hope they didn't tackle the c152 on 4788. I would hate for bsquared to have wasted his work

"They" would need quite a bit of hardware to scoop me on this one

.
I've had 40+ cores on it since yesterday, noon. Should be done sieving by this evening.

EdH · 2011-12-29, 20:59

I have had some of my sequences advanced by whomever, but I must disagree with locks for such. In viewing from the standpoint of the db, one would not want to inhibit access that would advance the whole.

The question still remains as to whether this is on purpose. I therefore do support a note of some kind on the sequence pages, not only for aliquot, but for any that have coordinated efforts.

I do support a very temporary lock on the issuance of composites. Not so that the db would not accept factors from anyone, but that the db would not issue the same composite within, say 24 hours. However, having said that, many of my machines, when running yoyo's script, fail to find factors due to memory allocation. These composites would not be reissued for a day.

Andi47 · 2011-12-29, 21:11

Quote:

Originally Posted by EdH

I do support a very temporary lock on the issuance of composites. Not so that the db would not accept factors from anyone, but that the db would not issue the same composite within, say 24 hours. However, having said that, many of my machines, when running yoyo's script, fail to find factors due to memory allocation. These composites would not be reissued for a day.

You mean, that after one of the elves asks for a bunch of random composites to factor, the same composites would not be random-issued to another elf during the next 24h? This would be a great idea (to avoid toestepping among the elves); maybe even better than a reservation system. Maybe composites which are entered / found when .elf-file type data (i.e. manually entered aliquot sequence factors) is submitted, should get this temporary lock too. (of course, this would only avoid "random-toestepping", so for avoiding "unknowingly not-on-purpose-"poaching"" a kind of message at the sequence-pages would be appreciated.)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Broken aliquot sequences	fivemack	FactorDB	46	2021-02-21 10:46
Broken aliquot sequences	schickel	FactorDB	18	2013-06-12 16:09
A new theorem about aliquot sequences	garambois	Aliquot Sequences	34	2012-06-10 21:53
All the open-ended Aliquot sequences.	Greebley	Aliquot Sequences	14	2011-03-15 00:13
New article on aliquot sequences	schickel	mersennewiki	0	2008-12-30 07:07

2011-12-28, 03:51	#12
LaurV Romulan Interpreter "name field" Jun 2011 Thailand 10,273 Posts	I agree (and vote for) a note on the DB page. We have nothing to lose from it. It happened to me many times that DB elves or someone else factored "easy" terms in my sequences, and few times even "hard" cofactors. I got the habit to avoid reporting the factors when they were "easy" to factor, they have less then 105 digits, the sequence got a downguide or downdriver, etc. The elves will not tease a big C125-C145 that fails the ecm, so they will spent their time more useful on something else. If you report small cofactors (under C105), then give it a time of few days before you start working on it. Do something else in this time. That is why I reserved 15 sequences. I won't work all of them in the same time. For example the elves factored few cofactors under C105 on my sequence 181428, or 865752. Be my guest! Very well done! :D This way, I don't duplicate their efforts and they don't duplicate mine (well, there are exceptions). I check the sequences with small cofactors from time to time and give them "pushes" when they seem stuck (usually ecm fail to crack it, and the elves lost their interest, that is most of the cofactors that need "a push" have some ECM already done on them - check factordb "scan" button - so I can directly schedule them for siqs/nfs after few ecm curves with high bounds). Of course, I get angry when the elves are faster then me. That is "human". I know for example that bchaffin elves seems to have preferences for sequences with downdrivers. :D... Or yoyo's elves will not check for reservations, but only for the size of the composite. You can be sure that any C100 or lower that you report could be "lucky" sooner or later to be grabbed by such an elf. If you report it, then give it few hours/days, and work something harder in that time (which was alredy "discarded" by the elves after a couple of ecm curves, or it have too many digits to ever be considered by them).

2011-12-28, 09:59	#13
yoyo Oct 2006 Berlin, Germany 659 Posts	The Boinc YAFU project just fetches composites from the factordb which are (currently) in the range from C80-C106 and send them as Boinc workunits to the Boinc elves. I suggested already in some other thread, that we should introduce a possibilit to reserve a composite for 2 days. After 2 days the resrvation is removed automatic. Than we can adapt our scripts and tools to fetch only unreserved composites and reserve them. Maybe I should chat with syd about it. yoyo

2011-12-29, 19:29	#18
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 19·23² Posts	Just think it through well if you will be installing locks. "If there's a gun on the wall in the first act, by the third act it will be used." Example: the perpetrator will go ahead of you for a few iterations and lock a future composite for herself (yes, I am being PC ). Instead, get and run another sequence and watch the "poached" one, -- and get the credit if it terminates. What's wrong with that?

2011-12-29, 20:14	#19
schickel "Frank <^>" Dec 2004 CDP Janesville 2·1,061 Posts	Well, I hope they didn't tackle the c152 on 4788. I would hate for bsquared to have wasted his work

2011-12-29, 20:59	#21
EdH "Ed Hall" Dec 2009 Adirondack Mtns 1010001101001₂ Posts	I have had some of my sequences advanced by whomever, but I must disagree with locks for such. In viewing from the standpoint of the db, one would not want to inhibit access that would advance the whole. The question still remains as to whether this is on purpose. I therefore do support a note of some kind on the sequence pages, not only for aliquot, but for any that have coordinated efforts. I do support a very temporary lock on the issuance of composites. Not so that the db would not accept factors from anyone, but that the db would not issue the same composite within, say 24 hours. However, having said that, many of my machines, when running yoyo's script, fail to find factors due to memory allocation. These composites would not be reissued for a day.