RSA-640

[ET_]

Originally Posted by Mystwalker

When I first read this, I thought "I don't think my skills fit. " - but then it came to my mind I don't even know what skills you need...

me, too... I'd like to help for such a project and still going on learning something more.

Luigi

[xilman]

Originally Posted by ET_

me, too... I'd like to help for such a project and still going on learning something more.

Luigi

Thanks for the offers. When Richard (aka Wacky) returns from his trip we will try to put together a detailed job specification, assuming the material below hasn't discouraged you.

It will certainly involve being able to write portable C/C++, knowledge of algorithms and data structures suitable for processing very large amounts of data in pseudo-real time (i.e. you can't wait until sieving is complete before starting work on data analysis!) on relatively ordinary computers. For instance, you can assume, say a hundred gigabytes of disk space, hundreds of megabytes of available memory and a cpu with a clockspeed of a couple of gigaHertz. You can not assume several terabytes of disk, multiple gigabytes of memory and a 64-way Itanium multiproc.

You will need to know what an XML parser looks like, or be able to find out quite quickly. You will need to be able either to implement a database to manage all the data and produce reports from it, or to know how to persuade an existing database to do it for you.

It's not likely that one person will do all of this, of course, so you will need to be able to work with other volunteers to create solutions between you.

There are certainly some challenges available for anyone willing to take them on. Unfortunately, and for reasons already outlined, the rewards are likely to be limited to respect from the community and the satisfaction of having done a good job. Oh, and bragging rights as well. The last may be a good thing to have on your CV.


Paul

[Minty]

Originally Posted by xilman

Earlier I wrote: By far the best boost we could get is more skilled labour. Serious offers only, please.
That was meant to be taken seriously.

And it was taken seriously - much earlier (in fact in my opening post!) i said -

Originally Posted by Minty

it was mentioned that there was a problem with manpower on the NFSNet side

so this is nothing new

It was pretty obvious that there was no interest in attacking RSA-640 from early on in the thread (even though 1kbit factorisations are romanced about frequently), but the reasons for this were not really obvious - sure, we hear and believe that the most necessary input is manpower, but we didn't know why and I for one was interested (to further my and everyone's knowledge). Without asking questions re the project (not much to read on such matters), how can we learn - after all, we're not telepathic!!!
Having picked up on the low level browbeating above (as others have), I am not enthused to help NFSNet in any way.

[jinydu]

Originally Posted by xilman

say a hundred gigabytes of disk space

That seems to be a lot to ask for. Most of the PCs I've seen (including my new, top-of-the-line notebook) still have much less than 100GB, and even for the computers that have more, 100GB seriously cuts into what a user can store on his computer.

To get more people to join the project, I think 10GB would probably be the maximum. But the 100GB version would be good for the most dedicated crunchers.

[BotXXX]

Originally Posted by jinydu

That seems to be a lot to ask for. Most of the PCs I've seen (including my new, top-of-the-line notebook) still have much less than 100GB, and even for the computers that have more, 100GB seriously cuts into what a user can store on his computer.

To get more people to join the project, I think 10GB would probably be the maximum. But the 100GB version would be good for the most dedicated crunchers.

The 100GB's seems to me to be for the server part. Dealing with all the lines brought in by the clients. The client it self sould not take more than 100~200 MB of harddisk space anything more will surethanly frighten other people. Ofcourse the math intressted will come, the money hunters (if there will be a cash reward) will be there but the normal user (which are needed in large numbers) will be scared away by to high space requirements.
At least that would be my vision.

[xilman]

Originally Posted by Minty

And it was taken seriously - much earlier (in fact in my opening post!) i said -

so this is nothing new

It was pretty obvious that there was no interest in attacking RSA-640 from early on in the thread (even though 1kbit factorisations are romanced about frequently), but the reasons for this were not really obvious - sure, we hear and believe that the most necessary input is manpower, but we didn't know why and I for one was interested (to further my and everyone's knowledge). Without asking questions re the project (not much to read on such matters), how can we learn - after all, we're not telepathic!!!
Having picked up on the low level browbeating above (as others have), I am not enthused to help NFSNet in any way.

Hey, relax. I know you were taking it seriously, and it seemed likely that others would also take it seriously.

That's why I spent time and effort explaining in detail the present situation and what the requirements may be for improving it. Questions were posted and I attempted to answer them. I answered the questions about finance with my honest belief. If you don't like my answers and disagree with them, so be it. All I can say is that I now have well over ten years experience of earning prize money from factoring RSA-challenge numbers and the cash has always been much more trouble than it's worth.

Until we can get the sievers upgraded to more than 30 or 32-bit working and the back-office functions greatly improved, it is not very practical to perform a 640-bit GNFS factorization.

As for the kilobit factorization you mention, I and one or two others have been working on that for a year or two. We are even further from running a 1024-bit SNFS than a 640-bit GNFS but that doesn't mean preparations can't be made. We can evaluate requirements and we can run ECM on candidates, for example. It would be senseless to use SNFS to discover a 50-digit factor when that would be much more easily found by ECM.

The front runner for a kilobit SNFS is R311, also known as 10,311-.c311 and (10^311-1)/9
Peter Montgomery and I spent about equal efforts running ECM first to the p45 level and then to p50. My resources are now somewhat less so I dropped out for the time being, but Peter and Bruce Dodson are carrying on with ECM searches optimized for p53, p55 and p60.


Paul

[xilman]

Originally Posted by BotXXX

The 100GB's seems to me to be for the server part. Dealing with all the lines brought in by the clients. The client it self sould not take more than 100~200 MB of harddisk space anything more will surethanly frighten other people. Ofcourse the math intressted will come, the money hunters (if there will be a cash reward) will be there but the normal user (which are needed in large numbers) will be scared away by to high space requirements.
At least that would be my vision.

Quite right. The topic under discussion was the need to improve the back-office functions. I apologise if this was unclear and I misled you.

The clients will, necessarily, be running on much less impressive machines. I would expect a 640-bit GNFS siever to take perhaps 200M disk (I doubt that the factor bases will fit in 100M, though we may be able to work with binary and/or compressed format) and 100M to 150M active virtual memory. Useful progress could be made on processors with clockspeeds well under a gigaHertz. In summary: with the exception of the memory requirements perhaps, most anything shipped in the last four years or so.

For a back-office system the requirements given are relatively modest. My home box, purchased 3.5 years ago, has a 60G disk, 768M RAM and a 1.3GHz Athlon. It was a chunky machine back then but is very modest now. It's hard to buy a new machine with a disk that small. At the moment, my machine has several copies of the data for both 7_233P and 6_257P loaded on it and still has 9G free space. It is also running the 7_233P matrix as I type, taking up 500M of that memory. Another day and it should finish, with factors appearing shortly afterwards with luck.

Paul

[xilman]

Originally Posted by xilman

As for the kilobit factorization you mention, I and one or two others have been working on that for a year or two. We are even further from running a 1024-bit SNFS than a 640-bit GNFS but that doesn't mean preparations can't be made. We can evaluate requirements and we can run ECM on candidates, for example. It would be senseless to use SNFS to discover a 50-digit factor when that would be much more easily found by ECM.

I should have added: we already know that the factors of RSA-640 will not be found by ECM except by the most astounding good fortune, so there is no point in looking for them. We know nothing about the factors of R311, other than that they are extremely likely to have at least 45 digits and probably more than 50, so intensive ECM effort is worthwhile.

Paul

[xilman]

Originally Posted by BotXXX

The 100GB's seems to me to be for the server part. Dealing with all the lines brought in by the clients.

Something else that I may have not explained very well: people writing the software do not need to have such a machine themselves!

Software can be developed and, to a large extent, tested on much less impressive machines. You can assume that someone else will provide the big iron for the back-office behind a full-scale factorization. All the raw data for two big projects (2_811M and 3_491P) is preserved, as is that for many smaller ones, so we can test any new implentation without having to sieve another and work on live data.

The system requirements are met easily by Wacky's G5 Macintosh and by the system I intend to get sometime in the next few weeks to replace my now elderly Athlon. Memory is now fairly cheap in 1-4G quantities, 200G disks don't cost that much any more and it's becoming quite difficult to buy processors with clock speeds much under 2GHz.


Paul

[ET_]

I'd like to help.

I've always wished to actively partecipate to a long running distributed math project and improve both my math and administrative (algorithmical) knowledge: after seven years of "passive" colaboration through running clients I think that practice may help more than theory.

I can offer disk space (ready to buy a 100 GB dedicated hard disk), some C and PHP programming capabilities, administration of a database and a network of four machines summing up to 7.5 GHz actually devoted to GIMPS and other distributed math projects. I may also ask for a static IP, and have web space with PHP and MySQL access.

Finally, I had experience of translating and mirroring sites and documentation into Italian and a deep passion for Number Theory

Luigi

[Minty]

Thanks for all the great info

Originally Posted by xilman

I would expect a 640-bit GNFS siever to take perhaps 200M disk (I doubt that the factor bases will fit in 100M, though we may be able to work with binary and/or compressed format) and 100M to 150M active virtual memory.

Wow! I didn't foresee these numbers!