[RC5] stop running the client?

Basil A. Daoust basildaoust at home.com
Tue Dec 14 12:50:39 EST 1999


This article which I can not read and have to assume makes a couple assumptions.
   1) How long the project will run.
      - This is not valid in rc5 as we don't know when we will find the key.
        Though someone with time could calculate a probability rating and time
        line for us...
     
   2) Fixed hardware
      - Assume you start running the test with the 1,000,000 computer and not
add
        anything or move the process to a new computer prior to finishing the
task

      Here we add new hardware daily, and loose hardware daily.  We don't have a 
      fixed amount of money to BUY anything.  Well maybe Dnet does but as long
as they
      can keep serving keys we can all keep cracking.

   3) We can upgrade the software on the fly.

The article I assume is using the theory that computers speed doubles every 18
months.
So if it takes 36 months for a project to complete today, wait 18 months buy a
computer
twice as fast for a bit less and finish at the same time.  Seems simple.  Sounds
reasonable for a program that fits the above profile.  Or maybe by 4+ of the
same computers and finish even faster.

Give your programers an extra 18 months to write faster/stabler code wouldn't
hurt any
project either :)  But don't tell them.

Basil

Stephen Berg wrote:
> 
> On Mon, 13 Dec 1999 00:12:39 -0600, Ryan Malayter wrote:
> 
> >The premise of this paper is that, for long-running calculations (like the Rc5 crack) it is often better to WAIT to start it on newer hadrware, beacuse you'll actually finish the task sooner.
> >
> >This is because of Moore's law. The author's calculations indicate that if any task will take longer than 26 months to complete on current hardware, you're better off waiting a wile and running it on new hardware.
> 
> But what if the correct key just got handed out to a client, that key
> block is going to be loaded into the client in say 20 minutes and
> processed in 10 minutes give or take a processor generation.  In a
> few hours the out buffer of that client will get full, it'll send the
> results with a possible hit flag to the proxy/key server and by
> tomorrow afternoon someone will be $1000 richer.  For something that
> we don't know the time it'll take I, (in my nowhere near knowledgable
> opinion on matters like this), would suspect that if we can start
> crunching now we might as well.  The odds may be against finding the
> solution in the first 20 or 30 percent, but not impossible.
> 
> I've seen similar articles saying we shouldn't bother trying to go to
> the nearest star until we can travel faster than the speed of light.
> The theory being that if we launch a long term mission now, it'll
> still be chugging along while technology builds a newer faster way to
> travel and later missions could actually arrive at that star *before*
> the first one arrives.  But in that case we pretty much know how long
> it'll take to finish with current technology.
> 
> If RC5 had been started on 8088's or 80286's back in the early 1980's
> could they have already finished RC5-64??  Maybe not without the
> internet access most of us enjoy today but I'd guess that the raw
> processor power was out there to at least start the job.  In the 20
> years or whatever we've had PC's around they could have checked a
> good 10 or 15% of the keyspace that we wouldn't have had to check
> now.
> 
> This is all a layman's opinion, those of you with Computer,
> Statistics or Math degrees can probably shoot most of this full of
> holes.
> 
>                             Stephen Berg
> //-    USAF Instructor  -/-  Reluctant NT User -/- Web Designer    -//
> //-                 Home = sberg at mississippi.com                   -//
> //-               Work = berg.stephen at keesler.af.mil               -//
> //-     http://iceberg.3c0x1.com/   -/-   http://www.3c0x1.com     -//
> 
> --
> To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
> rc5-digest subscribers replace rc5 with rc5-digest

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list