Mersenne Digest           Thursday, 11 March 1999      Volume 01 : Number 528


----------------------------------------------------------------------

From: Conrad Curry <[EMAIL PROTECTED]>
Date: Wed, 10 Mar 1999 15:38:52 -0600 (EST)
Subject: Re: Mersenne: VME claim

> With regards to the claim made by VME, Brian Beesley and I asked them to
> produce a factor of M(727). They did not come up with a factor. Instead
> they came up with the following (mass) reply which I leave to everyone's
> own thoughts. They also attached a letter in .gif format which can be
> viewed at http://home.wxs.nl/~tha/Mersenne/endorse.gif

  Before anyone believes anything from this company I would
strongly suggest reading the snake oil FAQ.  One version is at
http://www.cs.uu.nl/wais/html/na-dir/cryptography-faq/snake-oil.html

  If a strong pseudoprime test identifies a number as composite it
is definitely composite, otherwise it is a probable prime.  The chance
that it returns prime when it is actually composite is bounded by 1/4
for each base.  If a 100 bases are used then the chance of error is
(1/4)^100 = 6.2*10^-61.  It would be more likely the calculation was
affected by cosmic radiation.

  A probablistic primality test could be as fast as the timings given
by meganet on their deterministic primality test.  Their claim of 5
minutes for a sparc II workstation for a probabilistic primality test
of a 1000 bit number is absurd.

  In Dr. Milstein's endorsement of this primality test, he states "I
applied the assertions of the paper to a number of non-trivial values."
and "I did not develop rigorous proofs, but I did recast the proposed
techniques within a sound mathematical framework ... I also believe
that rigorous proofs are less important than validating the performance
of the algorithm."

  I emailed Dr. Milstein and asked if he has not rigorously proved this
then is it possible that it could a probabilisitic primality test.  This
is his  response:

"The answer to your question is NO, I've formulated the raw material
given to me by Meganet as a clear mathematical representation, i.e.
Lemma, Theorem, Corollary etc. I've proven some results and 'convince
myself' after a serious analysis that there is merit to the claim of 'a
deterministic algorithm'. Moreover, I've checked specific cases were
other techniques identified pseudo primes as primes or skipped all
together a specific prime. My first choice at this time, is to exhibit a
working algorithm rather than generating an analytical proof (I might do
it in the future)."

  Dr. Milstein states that there is "NO" possibility that this is a
probabilistic primality test, despite having no rigorous proof.  I don't
know what technique would identify a pseudoprime as prime.  I have some
doubts as to Dr. Milstein's ability or bias.

  In my opinion Meganet is making fraudelent claims of its software.
Why would anyone trust an unproven unpublished algorithm especially for
cryptography and primality testing?

  Meganet has sent me four emails about their claims, despite responding
with complaints they believe I had some interest.  Others on this list
have also received their emails which leads me to believe they may have
taken addresses from the mail archive.  So I have removed the zip files
from the web page.

  From a traceroute Sprintlink appears to be the upstream provider of
Meganet.  I have sent three complaints to [EMAIL PROTECTED]

________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: Jud McCranie <[EMAIL PROTECTED]>
Date: Wed, 10 Mar 1999 16:50:14 -0500
Subject: Re: Mersenne: Milstein

At 01:52 PM 3/10/99 -0500, Joth Tupper wrote:
>Thanks for sharing the bib search results.  There seems to be about a 10
>year gap between the
>2 sets of papers as well as some shift in field.
>
>The linear algebra papers 

My guess is that they are from his PhD theses, and perhaps Moshe Goldberg
was the advisor.

________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: "Ernst W. Mayer" <[EMAIL PROTECTED]>
Date: Wed, 10 Mar 1999 19:23:14 -0500
Subject: Mersenne: Re: buy 1 sphere, get the 2nd free!  

Joth Tupper writes:

>the Bolzano-Tarski theorem proved (what, back in the 1920's?) 
>that you could cut a solid 3D sphere into finitely many chunks,
>then rearrange the chunks to make another solid (no holes or gaps)
>3D sphere with _twice_ the volume.  Pretty spooky, I always felt.

I don't know about spooky, but 'twould seem to violate conservation
of mass (or mass/energy, if you're a postmodern relativist :),
'twouldn't it?

(BTW, the word "'twouldn't" has NOT been certified as correct
English by Professor J. Milstein - something about waiting for the
check to clear...)

Now back to our regular programming.

- -Ernst
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: "Leslie Burrow" <[EMAIL PROTECTED]>
Date: Thu, 11 Mar 1999 01:22:41 -0000
Subject: Re: Mersenne: Milstein

I was not the only one to run a check on Prof. Milstein on the web then.

It seemed odd that the only three recent mentions of Jaime Milstein are in
the field of linear algebra with the same co-author, Thomas L Moeller.

One wonders who the other "top ranking " mathematician that he submitted the
theory to could have been. <BG>

Les

________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: Marc Getty <[EMAIL PROTECTED]>
Date: Wed, 10 Mar 1999 20:23:15 -0500
Subject: Mersenne: Abolish PrimeNet? - Top Producers Chances - Aging Machines - Rate  
Rankings

Abolish PrimeNet?
- -----------------

> I guess we could do that by getting rid of primenet ;-)) if it was all done
> by email and manually entering the exponents on each machine (like in the
> old days) then I can still manage my 9 machines. Would you still run over
> 300? ;-)

Not a chance in hell. I spend 1 hour a week working with my machines,
thetas
it, no more. I spend that 1 hour a week on late Friday afternoons making
sure all the machines are up and running for the weekend. I can not
afford
to spend any more time then this. This is a good point, PrimeNet has
made it
incredibly more easy to run large setups like mine. Without PrimeNet I
would probably only run my fastest machines and the rest would have to 
simply run screen savers.

I know your comment was simply in jest, or a rhetorical comment but it
is my opinion that PrimeNet is the best thing to come around since
Prime95.
I ran Prime95 on my personal machine at home back before PrimeNet, and 
found the e-mail system to be quite cumbersome and left the GIMPS
effort.
I did not even look at the GIMPS site for about a year after this. But
when I did come back, last summer, I was so impressed with PrimeNet that
I gave it another shot. I ran it on my new at the time PII-350 and liked
it enough that I installed it on 200 of my 285 machines that September.
Without it I would probably still be running screen savers. Thanks
Scott!

Top Producers Chances
- ---------------------

> eh? if i buy 10,000 tickets and you buy one, then I have 10000 times more
> chances than you. OK that is still only 1 in 1400 (in the uk) as opposed to
> 1 in 14,000,000 but I know which I would choose....

FYI: Here in Pennsylvania in the US, our lottery odds are about 70
million to
one. My point is that everyone still has minuscule odds of winning. You
would
have a 1 in 7000 chance (0.14%) in winning vs my 1 in 70,000,000
(0.00014%),
not that great of odds for either of us, chances are someone else will
win.

Lets say that I bought 910 tickets (the equivalent number of P90s I
have)
and Stephen Wood bought 600 tickets (a guess at the equivalent number of
P90s
he has). Lets also say that the odds are 1 in 25,000 (an estimate as to
the
number of P90s equivalents participating in GIMPS). Overwhelmingly the
chances
are that someone else, with fewer machines, will find the next Mersenne 
Prime. Together, as the two leading producers in GIMPS we only have a 6%
chance of finding the next Mersenne Prime. Now granted these odds beat
the
pants off the odds of winning Pennsylvania's Lotto, but the fact remains
that
chances are someone else will be the winner either way.

Aging Machines
- --------------

> Aging machines - 97 (was it?) PII-300's I can think of a lot of corporates
> in the uk who would look in envy at your hardware setup and consider it
> pretty much state of the art....

And it could be 5 years before they are replaced. My lab of 48 PPro 200s
are
now old enough that they are on the schedule to be upgraded in the FY
2001-
2002. By this time they will be positively ancient. Yes, I realize that
many, perhaps most, envy at 97 PII-300 machines, but my point is that my
machines are only fast now, and will be slow soon. It may be the year
2003
or 2005 before the PII-300's are replaced. By this time someone with 50
PentiumIX 9 GHz machines will blow past me in the rankings like I am
standing
still. My 300 machines will not hold a candle to their 50 machines.

Rate Rankings
- -------------

> A good idea, how about we split the rankings into separate lists judged by
> the number of hours per day produced...
>   <50   <500   <5000

I think this is a poor idea. Not because I am currently on top, but what
purpose does this have? Don't get me wrong, there is nothing wrong with
yet
another page listing Top Producers, by rate rather then years produced.
There is nothing wrong with the spirit of what you are proposing, but
statistically there is no way to have the little guy at the top of any
list,
at least as far as I can tell.

A person with a single brand new PIII-500 will, after turning in one
result,
rocket into the middle of the <500 list as you propose. Does this make
this
person with a new hot rod of a computer a member of the multiple
machines at
home list along with the person with a PII-233, a Mac, a 486 and a
Laptop? Or
does this make this person a single machine owner that is bumped "down"
onto
a different list, one with people with one machine? Early on in my GIMPS
participation I corresponded with an engineer in the midwest who was
running
a whole bunch of 486/33's and was quite happy doing so. This makes him a
member of the <50 hours a day list, but he still has a bunch of
machines.
The rate of a participant in GIMPS can not display the number of
machines
they have producing that rate. One could have a single PIII-500, or an
entire truckload of 5 year old machines and still produce the same rate.

Other top producers lists, whether sorted by rate, by number of machines
or by rate per machine could segment the GIMPS effort even further.
Pages
representing these statistics would be quite informative, but none of
them
would allow the "little guy" to rise to the top of any of these lists. 
Only individuals/teams with large numbers of machines, or
individuals/teams
with the absolute newest bleeding edge machines would ever rise to the
top
of any top producers list.


Marc Getty  -  [EMAIL PROTECTED]  -  [EMAIL PROTECTED]  - ICQ: 12916278
  http://www.getty.net  -  http://www.vwthing.org  Work: 215-204-3291
           http://etc.temple.edu/                  Home: 215-322-8363
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: "Leslie Burrow" <[EMAIL PROTECTED]>
Date: Thu, 11 Mar 1999 01:34:40 -0000
Subject: Re: Mersenne: [Mersenne] Celeron

- -----Original Message-----
From: Jud McCranie <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED] <[EMAIL PROTECTED]>
Date: 10 March 1999 23:55
Subject: Mersenne: [Mersenne] Celeron

>How does the current Celeron stack up against a P-II for this work?
>
>
see my reply to Aaron Blosser about mid-January for comparison with a Xeon

currently LLtesting 6989963 getting 0.198secs per iteration (on a good day
with a following wind) with no other processes running .Except of course the
operating system which cannot be ignored as it's Win98.(I'm working on a
Linux set-up but don't have the experience or expertise to make the
transition easy; but it is fun learning something new every day)

Les

________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: "Aaron Blosser" <[EMAIL PROTECTED]>
Date: Wed, 10 Mar 1999 21:15:43 -0700
Subject: RE: Mersenne: [Mersenne] Celeron

> From: Jud McCranie
> Subject: Mersenne: [Mersenne] Celeron

> How does the current Celeron stack up against a P-II for this work?

I would speculate that it would be a tad bit slower:

1) The newer Celerons have a 128k L2 cache, less than the 256K L2 cache
found on the PII

2) Even though the Celeron L2 cache runs at CPU speed (rather than CPU/2
speed like the PII), it's smaller size negates any improvement.

What I'd really like to see are comparisons between Xeon's running at the
same speed but with different L2 sizes, including the 2MB L2 cache model.
But since those are REAL expensive, I don't expect anyone has any of those?

Aaron

________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: "Curtis (Jewell) Whalen" <[EMAIL PROTECTED]>
Date: Wed, 10 Mar 1999 23:01:08 -0600
Subject: Re: Mersenne: Mersenne Machine & Single Floppy LL Tester

>From: Marc Getty <[EMAIL PROTECTED]>
>Date: Wed, 10 Mar 1999 09:23:18 -0500
>Subject: Mersenne: Mersenne Machine & Single Floppy LL Tester

...
>Single Floppy LL Tester?
>========================
>
>What would be good to have is an all in one magic Prime95 bootable
>floppy disk. I would find it incredibly useful to have a bootable
>disk that runs LL tests w/o an Windows operation system on the hard
>disk of the machine.  ... If I
>could use these completely idle CPUs I could get perhaps 160 LL
>tests done that would not be done otherwise. I would also "burn-in"
>these machines at the same time. If anyone out there has a
>configuration like this already, please let me know!
>
>What I propose is to do one of the following:
>
>Method One
>- ----------
>1. Make a bootable disk using MS-DOS or MS-Windows95's DOS, the end
>user must provide this for software licensing reasons, GIMPS can't
>go around giving out MS software. Win95 B or Win98 bootable would
>be preferred so FAT32 partitions can be "seen".


I like this idea. Would FreeDOS work? I don't think it has Fat32 yet... but
GIMPS could certainly give THAT out.

...

>3. Being that rather large temporary files are created, often greater
>then a floppy disk, they can be temporarily redirected to the hard
>drive of the computer.


Assuming, of course, that your "floppy" is not a SuperDisk or a Zip.
If it is, why redirect?

>Method Two
>- ----------
>Use a linux bootable disk with mprime on it set to automatically
>load on boot, again only manual testing, with temporary files
>redirected to the hard drive of the computer. The linux kernel
>would have to be both FAT32 aware, because most new machines
>ship with FAT32 formatted hard drives now. Hell, if you are really
>good network support could also be built into this disk!
>
>I am not a linux guru by any means, but I'm pretty sure this is
>possible, and can then be freely distributed as a disk image.


I REALLY like this idea.

If this disk got distributed, maybe put LS-120 (SuperDisk) drivers on the disk
for an alternate temporary files location as well. I know Linux has them.

Or just make a big ramdrive!

- --Curtis

________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: lrwiman <[EMAIL PROTECTED]>
Date: Thu, 11 Mar 1999 00:16:37 -0500 (EST)
Subject: Mersenne: M(M3021377)

All,
How long would it take to run LL testing on M(M3021377) asuming that this
number was prime.  Could we complete it before the sun explodes?

Could the Litho-universe computer complete it before protons start to decay?
- -Lucas
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: Greg Hewgill <[EMAIL PROTECTED]>
Date: Wed, 10 Mar 1999 23:38:36 -0800
Subject: Re: Mersenne: Mersenne Machine & Single Floppy LL Tester

On Wed, Mar 10, 1999 at 11:01:08PM -0600, Curtis (Jewell) Whalen wrote:
> >Method Two
> >- ----------
> >Use a linux bootable disk with mprime on it set to automatically
> >load on boot, again only manual testing, with temporary files
>
> I REALLY like this idea.

I've actually done something like this. I have three machines, that have
essentially the following components:

- - Celeron (of varying overclocked speeds)
- - 440BX motherboard (for 100 MHz memory bus)
- - 32 MB PC100 RAM
- - NE2000 clone network adapter
- - floppy drive
- - case (not into wiring my own power supplies)

No keyboard, video, or hard drive. The floppy boots a Linux kernel which
mounts an NFS root file system from a networked P90 (the same P90 is the
file server for all three machines). The mprime files are stored on the
P90 - in fact, I only have one boot disk, I just move it from one machine
to another when I boot them. They talk to the primenet server through an
HTTP proxy, and are pretty much hands-off operation. Works really well.

Greg Hewgill
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

From: "John R Pierce" <[EMAIL PROTECTED]>
Date: Wed, 10 Mar 1999 23:52:58 -0800
Subject: Re: Mersenne: M(M3021377)

> All,
> How long would it take to run LL testing on M(M3021377) asuming that this
> number was prime.  Could we complete it before the sun explodes?

Oh, I suspect if we could build a computer with a bunch of 4 million bit wide
multipliers that ran at a few terahertz it would take significantly less time
than that.

- -jrp


________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm

------------------------------

End of Mersenne Digest V1 #528
******************************

Reply via email to