I have set up a deeply primitive version of the reservation system at

[url]http://www.chiark.greenend.org.uk/ucgi/~twomack/homcun.pl[/url]

Could people have a look at it, check in what they're doing, check how the program breaks, and tell me (tom@womack.net) anything they want added. There's an infinite amount of potential galvinising and gold-plating to do, I'd like to know where to start. Should I copy Kamada's idea of giving a short code when you reserve a number which you have to submit with a factor, to make work-poaching less of an issue?

At the moment it doesn't handle submission of factors; my temptation is that you simply submit a factor, and it does the GCDs to work out what the number was a factor of, before taking you to a submit-more-details form. There's an issue if the number splits as p1*p2*p3 and someone reserves p2*p3 between your submission of p1 and of p2, but I don't see that happening very often.

I have 6^239-5^239 about half-way done, but I noticed that Al Edwards has just reserved it. At this point, I do plan to finish it.

Yep, you're right. Reservation system definitely needed!

Greg

[QUOTE=fivemack;99364]tell me (tom@womack.net) anything they want added. There's an infinite amount of potential galvinising and gold-plating to do, I'd like to know where to start.[/QUOTE]When I had a look earlier today, there didn't seem to be a facility to unreserve a number. That would be useful, IMO.

Submission of factors, etc, would be nice (especially if the software also sends me a Cc:) but I don't personally see it as an essential feature straight away. The essentials are the ability to reserve a number, to unreserve one of your own numbers, and for an admin to forcibly unreserve a number that has been abandoned.

Unreservation, of course, will require a nonce to be issued to the reserver and the return of that nonce when the number is to be returned to the free list. You're already thinking along those lines anyway.


Paul

[QUOTE=frmky;99383]I have 6^239-5^239 about half-way done, but I noticed that Al Edwards has just reserved it. At this point, I do plan to finish it.

Yep, you're right. Reservation system definitely needed!

Greg[/QUOTE]You're way ahead of me: it's yours.

I usually claim one only after I have started sieving. I just put 2000 curves into this at B1=11e6, and had set up an ggnfs .poly file.

Yes: definitely could use an "unreserve" utility.

I have added an unreserve facility, you may have noticed Testy the Test Fairy reserving and unreserving some numbers. I've coded up two-thirds of a 'submit' facility, which I think probably is useful, but it's not ready for use yet, and it's nearly eleven on Sunday night after a week-long conference and a friend's wedding on Saturday, so I should probably sleep instead.

I'm tempted not to require a code for submission, just credit submissions of a factor by anybody to the person who's got it reserved. Should I have a text-field requesting information about the circumstances of the factorisation (in which case I'd better sort out sending emails to Paul), or just trust that factors will get submitted to Paul anyway?

Hi,

I also "reserved" those for which I most recently submitted results.

A couple odd bugs in the reservation system... In many cases, common factors haven't been eliminated, i.e., 5,2,253+ is listed as 10^253+4^253. Also, the SNFS difficulty isn't right in many cases, though I realize that it's not always trivial to find automatically. For example, 5,2,253+ had SNFS difficulty of 160.8, not 253, while 5,3,288+ has difficulty of 134.2, not 83.9.

Thanks for doing the work to put this system together!

Greg

Another result...

6,5,253-
23399715679732609185037453179199938393534804811895004040917823 (pp62)
56629592391441112515970389366375799362154413221434713125631953178284579124509 (pp77)

Greg

[QUOTE=frmky;99399]
Also, the SNFS difficulty isn't right in many cases, though I realize that it's not always trivial to find automatically. For example, 5,2,253+ had SNFS difficulty of 160.8, not 253, while 5,3,288+ has difficulty of 134.2, not 83.9.
[/QUOTE]

OK, I've worked out what's going wrong. I find the labels by writing a list of a^b mod N with between 90 and 270 digits and sorting; but 2^253 has less than 90 digits, so wasn't in the list. 4^253 was, so I picked up 5^253+2^253 as 10^253+4^253.

Is the factor 11 there actually usable by a reciprocal trick for this kind of number?

5,3,288+, my excuse is that I applied a (N-1)/N factor to the difficulty for all small primes dividing b rather than for the smallest, and forgot to apply only odd numbers in the + case.

I've corrected those in the code that generates the table of difficulties from comps.gz, merged into the live database, and removed the two already-submitted results that I know about (6,5,253- and 5,3,251+).

Would sort-by-SNFS-difficulty be useful?

[QUOTE=fivemack;99414]
Is the factor 11 there actually usable by a reciprocal trick for this kind of number?
[/QUOTE]

Yes, it is. So is a factor of 13.

[QUOTE=fivemack;99414]
Would sort-by-SNFS-difficulty be useful?[/QUOTE]

I think it would be. Also useful would be a listing of the full expansion of the number after it is reserved so I don't have to calculate it, and a link on the reservation page to Paul's main page.

When I know a factor has been found by ECMNET (by myself, for instance) would it be appropriate to "reserve" it so that no one else does, or should I leave it unreserved to be removed in the next update?

Thanks again!
Greg

There's still something odd going on with the SNFS difficulty. For example, 266 = 7*38, so 5^266+4^266 has difficulty of 159.4.

Greg

I've added the link to Paul's page and the contributions from 11 and 13 to SNFS difficulty.

If a factor's found by ECMNET, reserving the number is the right thing to do at the moment; I'll code up the remaining third of the submission routine tonight ... thorough testing rather required me to have a factor to work with.

If you follow the 'unreserve' link from one of your numbers, the number is written out in full in the URL and on the page you get, and there's no danger of unreserving the number by mistake because you have to enter the reservation code. I suppose the right answer is to add a link from the description to a page which displays the whole number and the snfs.poly input wanted by ggnfs, though that will be interesting to write and test; maybe by the end of the week?