On Mon, 13 Sep 1999 23:24:12 -0400, George wrote:
>Hi all,
>
>At 09:14 PM 9/13/99 -0400, Tom Goulet wrote:
>>I wish to offer constructive feedback.
>>If you must flame me, do it off-list. :)
>
>Constructive criticism is always welcome, although you will likely
>get some flames :)
No flames so far, at least in public :)
<snip>
>
>>The GIMPS clients does a lot of self-testing...is that really needed?
>
>Not any more. In the old days, the one-hour self-test was the only
>way prime95 was able to detect faulty machines. Now there is a test done
>on each iteration so that faulty machines are discovered eventually.
>The self-test/torture-test is now useful for people that want to stress
>test a new computer - and yes a surprising number of machines fail
>prime95's torture test even when all other programs run perfectly.
It has also been said in the past that the self test is to test the
dedication of the tester. Anyone who is prepared to wait a few hours
for the self-test should have the patience to wait a month for the
result. :)
>>The 'packet' sizes are very large...I have lost weeks of work due to
>>power failures and friends.
>
>The math algorithms are the cause of the large 'packet' sizes - and
>the problem is only going to get worse. It won't be long before a
>single exponent takes one to three months to test.
>
>Smaller packets would require using lots of bandwidth and server storage
>to store intermediate files. Also, the end result would only be as reliable
>as the buggiest machine that worked on the result. In other words, one
>overclocked CPU could pollute dozens of results with smaller work units.
>
If you want something that is shorter, you could always try factoring
smaller Mersennes. http://www.mersenne.org/ecm.htm has the details and
you can do it with the existing Mersenne client. But don't try it for
exponents between 5000 & 6000 because they're all mine!!! <maniacal
laugh>
>
>>So anyways, my comments can be summed up in that it isn't easy for the newbie
>>(even the very nerdy types like myself) to find their way around the gimps
>>project and become efficient contributers.
>>Now, distributed.net (yes, I hear some moans) has a very slick interface, and
>>helpful documentation and navigable websites and easy to understand overall
>>structure.
>
>This is my fault. I believe d.net has many people working on the client
>and server. GIMPS is basically run by myself and Scott. My interests are
>in the math and optimization aspects of the program. Thus, given a choice
>between making the web pages slicker, adding a nice setup program, writing
>a hefty help file, or speeding up and making more use of the P-1 factoring
>code, then I'll choose the latter simply because I enjoy that more.
>
>That said, I do take enough pride in my work to make sure bugs are fixed
>promptly and the user interface gets a few new features each release.
>Actually, many find it very easy to install and run but get confused when
>trying to figure out what the program is doing.
>
If it's any consolation George, I had no trouble getting the client
working. That was version 13, I think. (This makes me feel old, but
someone will probably say they remember version 1 as if it was
yesterday)
Many of the problems of new users (this is not directed at Tom) are
answered in the readme file. It's not as flash as a searchable help
facility but it does its job.
>>Hopefully this explains to some of you why the gimps is not growing very
>fast.
>
>Believe it or not, growing GIMPS at the fastest possible rate is not
>my personal goal (Scott and many other GIMPS members probably wish it was!)
I think the slow growth has more to do with:
1. It takes over a month to get a result, at least on my PC.
2. 99,999 times out of 100,000, this is negative
3. Mathematics doesn't have the romantic appeal of say SETI or
code-cracking, to most people.
4. People don't know about it. There's a surge of new recruits every
time we get more publicity but it then dies down.
>I find it more rewarding when a GIMPS member has always been interested
>in primes and number theory or when GIMPS inspires someone to discover
>the joy of recreational mathematics.
>
Certainly it's made me try to learn something about the background.
Now all I need is a decent explanation of the number field sieve. :)
--
Cheers
Steve Whitaker
31 new factors and counting
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
