Value, or otherwise, of running others' code

[xilman]

Originally Posted by bdodson

Uhm. Seems a bit sharp

It was intended to be sharp, AFAICT. Funny, though.

Paul

[michaf]

As a beside, if someone LIKES to do work on HP's, cunninghams, search for E.T., let it be so.

As long as the person doing the work is happy, all's fine :)
It's _their_ computer power after all.

And now be happy all, please :)

[R.D. Silverman]

[QUOTE=michaf;120002]As a beside, if someone LIKES to do work on HP's, cunninghams, search for E.T., let it be so.

As long as the person doing the work is happy, all's fine :)
It's _their_ computer power after all.

QUOTE]


It may be their computer, but it is someone else's code.

[Zeta-Flux]

Silverman,

Are you a code Nazi? Are you suggesting that we all build our computers from scratch because it would simply be wrong to use the chipset instructions put in place by someone else for our pointless projects?? No more using Mathematica for me!

[akruppa]

Originally Posted by R.D. Silverman

It may be their computer, but it is someone else's code.

I don't see how this would keep anyone from spending their cpu time on their particular pet project, so long as the license agreement of the code does not forbid it. No such restriction exists in the licenses of GGNFS, msieve, GMP-ECM, Alpertron's applets, Prime95 or any factoring software I'm aware of.

Alex

[R.D. Silverman]

Originally Posted by akruppa

I don't see how this would keep anyone from spending their cpu time on their particular pet project, so long as the license agreement of the code does not forbid it. No such restriction exists in the licenses of GGNFS, msieve, GMP-ECM, Alpertron's applets, Prime95 or any factoring software I'm aware of.

Alex

When I was first introduced to computational number theory I was
every bit as ignorant about it (well, almost; I did have a math degree, but
had not taken any number theory) as others in this forum.

I found it interesting. And as a result, I wanted to participate.
It never even *occured to me* to ask others for their executables
so I could participate. Instead, I went out and READ. I read papers.
I read textbooks on number theory. I read books on how to write good
numerical code. Then I went out *AND WROTE MY OWN CODE*.
Yes, I asked for help and advice from time to time.

Then, and *only then* could I take pride in making a contribution
in computational number theory.

And I have a right to be contemptuous of those who just want to
blindly run code written by others and then want to claim "credit" for
finding some result.

[retina]

Looks like I need to stop my prime95 and other software (including the OS) 'cause I didn't write it myself. Also, I need to turn off and sell my PC's 'cause I didn't make them myself. I had better turnoff the lights and other electrical appliances since I didn't design and build them myself. Oh yeah, my house was not built by me, that has to go as well. Plus my clothes, currency, comb, soap, car, etc.; all must be abandoned. We all have to start from zero and learn to make and do everything on our own else the Silverman fairy will come and nag us to death for using something we didn't make/build/write/think-of ourselves.

It's called a society, I do things you want and need and you do things I want and need. If we all did everything ourselves then we'd still be in the stone ages. Some people work in mines to keep your air-con/heater going to keep you cool/warm while you write code to give them, umm ...what, something. Anyhow, you can't expect everyone to have time to write/debug/test their own code. They have to get back to mine each day to keep the rest of us comfortable.

[retina]

Originally Posted by R.D. Silverman

<snip>

It would only take me a few hours work to convert my SNFS code to
a full-blown GNFS code except for one component: polynomial selection.
I could do some GNFS factorizations, but I would have to ask others to
find polynomials for me. If I had a full understanding of Brian Murphy's
thesis and *could* implement polynomial selection code (but simply
lack the time) I would feel better about asking others....But I don't.
And I don't feel right in asking others to do my thinking for me.

This is a good point but, you can also look at it in a different way. Two heads are better than one. You have contributed the SNFS code and someone else might want to contribute the factorization code. Thus together you can make something better then either of you alone could have had the time to achieve. This way the contributions are more like 1+1=3.

I don't see it as a matter of others doing your thinking for you. It is more like others contributing different parts to make a greater whole.

[R.D. Silverman]

Originally Posted by retina

This is a good point but, you can also look at it in a different way. Two heads are better than one. You have contributed the SNFS code and someone else might want to contribute the factorization code. Thus together you can make something better then either of you alone could have had the time to achieve. This way the contributions are more like 1+1=3.

I don't see it as a matter of others doing your thinking for you. It is more like others contributing different parts to make a greater whole.

This seems reasonable. Except too many people are not contributing
"parts". All they contribute is CPU time. And writing a "wrapper" that
simply calls GMP-ECM to do the work isn't really contributing a new
capability. One can write a simple little shell script or .bat file that
calls GMP-ECM with different parameters to first run P-1, then P+1
etc.

If one really wants to make a contribution, then take the GMP code and
find/implement a small improvement. Or make a small improvement to the
ECM code. Or improve the msieve code, etc. etc. This would have
value to others.

[xilman]

Originally Posted by R.D. Silverman

This seems reasonable. Except too many people are not contributing
"parts". All they contribute is CPU time. And writing a "wrapper" that
simply calls GMP-ECM to do the work isn't really contributing a new
capability. One can write a simple little shell script or .bat file that
calls GMP-ECM with different parameters to first run P-1, then P+1
etc.

If one really wants to make a contribution, then take the GMP code and
find/implement a small improvement. Or make a small improvement to the
ECM code. Or improve the msieve code, etc. etc. This would have
value to others.

As I threatened in the parent thread --- some feather-ruffling.

I rather like people contributing their CPU time to my projects. I'm a firm believer in getting other people to do my work for me. It frees up my resources to do other work I want to do.

There: I've probably offended both sides of this particular argument.

(As I've written on many previous occasions: anyone reading anything I post should ensure that their irony detector has been serviced recently. Make of that what you will.)


Paul

[ewmayer]

Flame and counterflame aside, I think the point Bob is trying to make is that these types of compute-some-in-itself-fairly-useless-thing projects only derive any real long-term value from inspiring people to learn the maths and computational skills needed to write the kind of code used by ... these types of compute-some-in-itself-fairly-useless-thing projects. So even taking that tack we end up with the snake eating its own tail, but the "out" is that the above types of skills will [hopefully] prove useful in a broader setting. For instance [using GIMPS as an example], extremely fast FFTs have uses far beyond searching for primes. Continually pushing the limits of how-large-we-can-factor gives us a good sense of the security of the various encryption schemes used in all modern e-commerce and e-communication. Writing and running state-of-the-art crunching code can be useful to computer architects, by way of revealing the bottlenecks in current microprocessor and memory architectures and showing what types of hardware improvements are likely to give the best bang for one's silicon buck.

The types of skill/time/motivation required to make a genuine contribution will inevitably be beyond that of most of the project participants, but it is important to never lose sight of the true value of these kinds of "recreational" projects, which is to push the state of the art in algorithmics and computation [both in-CPU and distributed], and hopefully inspire at least some non-negligible fraction of participants to make an effort to do what is needed to make a genuine contribution. If that makes Bob a "code Nazi," well, I guess it's time for me to pull on a pair of computational jackboots, too. Bob is just more blunt about the value of merely running a DC client than most DC project leaders can afford to be, if they want their project to grow its user base. Which brings me to a valid counter-point: for many of the people [not all, but many] who actually *are* doing the hard work needed to write world-class code for these projects, the prospect of having >> 1 people running their code does provide a very real incentive. Sure it's a bit vain, but 'tis human nature.