Another colossal waste of time?

[rogue]

I just saw this update at distributed.net. I wonder if the new project will be a colossal waste of computing resources like RC72.

[mdettweiler]

Originally Posted by rogue

I just saw this update at distributed.net. I wonder if the new project will be a colossal waste of computing resources like RC72.

Well, RC5-72 probably wouldn't have been a waste of time if the prize hadn't been abandoned. And Distributed.net would have had no way of knowing that the prize wouldn't last when they started it, so I guess it's not exactly their fault.

[rogue]

You must be kidding, right? As of today, they have been running the project for 1803 days and have only searched 0.452% of the keyspace. If the project continues at a fixed rate, it would take 1092 years to complete. Unfortunately that hasn't been the case. They still get new participants who do not seem to understand the futility of the search.

I remember when they started this project. At the time, I did some "back of the envelope" calculations and estimated that if they doubled their resources every two years that it would take almost 20 years to search the entire keyspace. Even if the prize were still available the computational cost will far exceed the value of the prize. The cost to run the dnet server would exceed the value of the prize.

What they have tried to do in the past is prove that various forms of encryption are not powerful enough. I think they were successful when looking for a 56-bit key and maybe for the 64-bit key, but the 72-bit key problem will not be solved without new technology. The only technology that is likely to solve this problem in a reasonable amount of time is a quantum computer and based upon some of the advances of the past few years, one might solve this problem before their brute force method solves it.

I personally think they should abandon the project and find something better to do with all of the computing resources that they have available to them. They are many worthy distributed projects out there and some have a much bigger payback.

[xilman]

Originally Posted by rogue

think they were successful when looking for a 56-bit key and maybe for the 64-bit key, but the 72-bit key problem will not be solved without new technology.

Personally, I think new insight (that is, new theoretical analysis and the implementation of new algorithms) is much the more likely way of finding the key.


Paul

[mdettweiler]

Originally Posted by rogue

You must be kidding, right? As of today, they have been running the project for 1803 days and have only searched 0.452% of the keyspace. If the project continues at a fixed rate, it would take 1092 years to complete. Unfortunately that hasn't been the case. They still get new participants who do not seem to understand the futility of the search.

I remember when they started this project. At the time, I did some "back of the envelope" calculations and estimated that if they doubled their resources every two years that it would take almost 20 years to search the entire keyspace. Even if the prize were still available the computational cost will far exceed the value of the prize. The cost to run the dnet server would exceed the value of the prize.

What they have tried to do in the past is prove that various forms of encryption are not powerful enough. I think they were successful when looking for a 56-bit key and maybe for the 64-bit key, but the 72-bit key problem will not be solved without new technology. The only technology that is likely to solve this problem in a reasonable amount of time is a quantum computer and based upon some of the advances of the past few years, one might solve this problem before their brute force method solves it.

I personally think they should abandon the project and find something better to do with all of the computing resources that they have available to them. They are many worthy distributed projects out there and some have a much bigger payback.

Okay, I see now. Yeah, I guess it is futile if they've only searched 0.452% of the keyspace! Probably the only chance of them breaking RC5-72 in any decent amount of time would be if the correct key happened to be near the beginning of the keyspace--and I guess the probability of that is enough to be able to say "it's uncrackable with current technology".

Thanks for correcting me on that! I didn't realize that they were only as far as (not) far as they are through the keyspace.

Let's hope that their new project will be more feasible, as well as useful too (so it's not a waste of time, no matter how feasible). Also let's hope it's something interesting--their membership has been declining recently, and an exciting new project is what they need to revitalize their membership.

[jasonp]

Originally Posted by rogue

I personally think they should abandon the project and find something better to do with all of the computing resources that they have available to them. They are many worthy distributed projects out there and some have a much bigger payback.

I talked to Decio via email, and d.net is in a real bind. All of their backend infrastructure is highly customized to the few distributed problems they've been working on, and while there is no shortage of interesting projects (I've suggested a few, go ahead and guess my preference based on the preponderance of my forum posts :) the expectation right now is that anyone who wants to use distributed.net has to manually integrate their applications with the existing infrastructure. So nobody is lining up to do so, and they don't have the manpower to do it themselves, and the project that was their main draw is now increasingly seen as irrelevant so membership is declining.

[jasong]

Originally Posted by jasonp

All of their backend infrastructure is highly customized to the few distributed problems they've been working on, and while there is no shortage of interesting projects (I've suggested a few, go ahead and guess my preference based on the preponderance of my forum posts :) the expectation right now is that anyone who wants to use distributed.net has to manually integrate their applications with the existing infrastructure.

Could you either post the general setup or give us a link? Better yet, tell us what made the setup optimal, so we can help come up with ideas.

[jasonp]

Originally Posted by jasong

Could you either post the general setup or give us a link? Better yet, tell us what made the setup optimal, so we can help come up with ideas.

I'm sorry, he didn't provide any details and I didn't ask for any. They do not have a usable external interface like BOINC does, and any project that wants to take advantage of d.net has an integration challenge ahead of them. Resolving this chicken-and-egg problem (no open interface -> no projects joining -> no evolution) is hopefully the big secret they're working on.

I should also clarify that our conversation was months ago, and I'm just as much in the dark as everyone else is about what they're planning currently