home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.crypt
- Path: sparky!uunet!cis.ohio-state.edu!magnus.acs.ohio-state.edu!usenet.ins.cwru.edu!agate!linus!linus.mitre.org!gauss!bs
- From: bs@gauss.mitre.org (Robert D. Silverman)
- Subject: Re: RSA-129 contest
- Message-ID: <1992Aug31.184355.16783@linus.mitre.org>
- Sender: news@linus.mitre.org (News Service)
- Nntp-Posting-Host: gauss.mitre.org
- Organization: Research Computer Facility, MITRE Corporation, Bedford, MA
- References: <19122.Aug2700.12.0192@virtualnews.nyu.edu> <1992Aug27.111639.16325@linus.mitre.org> <18112.Aug3004.30.1292@virtualnews.nyu.edu>
- Date: Mon, 31 Aug 1992 18:43:55 GMT
- Lines: 82
-
- In article <18112.Aug3004.30.1292@virtualnews.nyu.edu> brnstnd@nyu.edu (D. J. Bernstein) writes:
- :In article <1992Aug27.111639.16325@linus.mitre.org> bs@gauss.mitre.org (Robert D. Silverman) writes:
- :> :Why are you trying to discourage people from
- :> :entering the RSA-129 contest, Bob?
- :> Now this last accusation really is obnoxious, because it puts words
- :> in my mouth that I have never said.
- :
- :Bob, you stated that a 6-7 digit improvement in GNFS polynomials in this
- :range would result in only a 2x speedup. If this were true, it would put
- :a rather small cap on the possible benefits of the RSA-129 contest.
- :
- :It is not, in fact, true. It is wildly inaccurate: as I pointed out
- :before, such an improvement could easily gain an order of magnitude in
-
- Having done quite a few numbers with the special number field sieve,
- I have observed that adding between 25 and 30 decimal digits to the
- size of the number I am factoring [using a 5th degree polynomial]
- roughly doubles the run time. Adding 30 digits increases the norm
- on the integer side by about 6 digits, and does not change the algebraic
- side. (that is, as long as the polynomial degree stays the same and the
- coefficients are single digit. [true for the special NFS].)
- If the factor base sizes are well chosen, it therefore seems that
- a 6 decimal digit increase on either side of the congruence [but not both]
- roughly doubles the run time.
-
- :achieving the effect of lying to the public, even if you don't mean to
- :do so. I don't care whether it's obnoxious of me to point this out---I
- :can't sit by and watch you corrupt the views of people who haven't
- :learned about GNFS and can only go by what they hear from others.
- :
- :> All you are doing is gainsaying my data without posting data of your own.
- :
- :That's right. The bulk of data which I have is part of an article I'm
- :writing with Arjen Lenstra on our GNFS implementation. That data, like
- :your data, is not sufficient to conclude that GNFS will probably be
- :slower than MPQS at 129 digits. Yet you keep stating this conclusion.
- :
-
- I'll let others judge your reply for themselves.
-
- My data supports a crossover point of 135 digits. It might very well
- be conservative by 10 or so digits. However, I have never represented
- otherwise than that it is an estimate based upon extrapolations of other data.
-
- If you have data that proves me wrong, I would be interested in hearing
- about it.
-
- However your continued personal attacks upon me because you don't agree
- with my estimates are totally uncalled for.
-
- :My apologies. I mistakenly thought that you were talking about Dickman's
- :rho function, which (I can say with some assurance now that I've checked
- :my references) a few people attribute to Knuth and Trabb Pardo (because
- :Knuth and Trabb Pardo did publish some things about it).
- :
- :I guess you are referring to the logarithmic sum that Schroeppel
- :originally suggested for choosing multiples for the continued fraction
-
- Yes. A variation of this function can also be applied to NFS. As with
- QS its value can vary quite a bit for numbers of a fixed size, and it
- seems as with QS that the higher the value, the easier it is to factor
- a number with NFS. The variation can be substantial. [I'm sure you know
- all this; I'm not trying to be patronizing]
-
- :method. If so, I am astounded at your statement that it would be nice to
- :have such a function for the algebraic side of GNFS---everyone knows
- :what it is, namely the sum of log p times the (easily computable)
- :probability that a coprime pair a,b has N(a,b) divisible by p^d, over
- :all p and d.
-
- Sorry for my lack of precision. I should have said one that can
- be computed from the polynomial coefficients directly. Of course if
- you know which primes split, which ramify, which are inert, and which
- ones split completely, one can compute this. But this means computing
- the factor base for every potential polynomial. [which really isn't
- that big a deal, I agree, but it is more expensive than changing multipliers
- for QS]
- --
- Bob Silverman
- These are my opinions and not MITRE's.
- Mitre Corporation, Bedford, MA 01730
- "You can lead a horse's ass to knowledge, but you can't make him think"
-