Please downgrade your d.net clients to v2.8010 or before

[Eug]

so anyone know what the prob is with 8011?

 

[Poof]

It's COSM not cosmo..

If you're talking about Adam Beberg, he and his COSM group have been working with the Stanford folks on Folding@home.

 

[Engine]

I don't know that Beberg & co. have been really working directly with the Folding@home project. From what I could gather from the Folding@home website, they just used Cosm's sdk to develop the project. I don't think that the Cosm people had any direct work in the project.

 

[Poof]

I don't think that the Cosm people had any direct work in the project.

Engine - he most certainly has. He posts regularly on their discussion list, which sortof fascinated the hell out of me...

He's been working with them to get HTTP proxy support working via the COSM libraries and the folding client.

[EDIT: Here's one of his recent posting's:

From: beberg@m...
Date: Wed Feb 21, 2001 11:55pm
Subject: Re: Thinking of throwing in the towel


> Why not do RC5 that has a bazillion options,
> reliable keyservers, and everything that you ever wanted in a
client
> included? The only reason I can see is that it is useless.

Well since I wrote the original RC5 client... and the original RC5
stats too... I can tell you that compared to what Folding@h... does,
RC5 is a completely TRIVIAL project.

RC5 (or SETI for that matter) have no data dependencies, no time
limits on the data return, no server coordination is required, no
Fortran libraries or voodoo math to integrate in, no actually
_science_ is involved (just had to count keys), no verification,
visualisation and checking against real experimental data (xray etc)
is required... I could go on. This project really pushes the envelope
in every direction. In other words, it's a real b*tch to get
everything right.

So you'll excuse us while we work out the bugs. We're doing our best,
but you cannot fairly compare Folding@h... to distributed.net, or
SETI, or anything else.

- Adam L. "Duncan" Beberg
http://www.iit.edu/~beberg/
beberg@m...

 

[Engine]

Oh... cool. It's good to see that he's involved, then

 

[RC5Bri]

Thank goodness for my pproxy! I only have 1 client running 8012 at work, and my main computer at home. That isn't too bad, only need to downgrade 2 out of 12!

Thanks for the info,
RC5Bri

 

[otlg24]

Poof:

Very familiar with the exploits of both Mr. Beberg and the folding@home crew... I speak with Vijay@stanford on a psuedo regular basis.

Regards,

Steve

 

[boomerang]

Well, I guess I'm glad that everyone is taking this so well.

I wish I could say the same. With all the stats problems and now this, I'm starting to think about all the electricity I could save by abandoning this project.

It looks as though RC5 will never end. We'll be crunching and re-crunching and yet again re-crunching. All due to circumstances beyond our control.

I may upgrade to the pre-RC5 client.

 

[ZapZilla]

Switching and flushing all buffers...

Glad I set RC5=0, thus, (hopefully) avoiding the described bug.

<dagnabity spelling err>

 

[Poof]

Yeah Engine - alot of us are consoled that he's involved, considering the complexity of the project!

And Boomerang - take this to heart. Bezerkeley has been down now since around 3am PST (11am UTC) today. As of this posting,they're still down. Rumors are circulating that some contractor accidently cut the fiber bundle on campus killing their connection out to the world. If this is true, who knows how long it might take to get them back online and even if it were immediate, it would probably be on a much smaller backup circuit (fiber is a bit tricky to deal with as I well know). And at this point, they would be DDoSed to death.

I have Setiqueue going with about 6 days worth of WUs stored up (been there done that ) for my couple of machines that are crunching.

At least I'm secure in the knowledge that my packets are getting flushed through dnet (whether they get lost in the stats or not)...

Stats problems aside, I don't really recall any network-related problems from dnet.

 

[Russ]

I'm still not clear on a couple of things. First:



<< Causing the RC5 block to be finished right away. >>



Does this statement mean that the block totals achieved were artificially high? And, if so, was this bug something that could be exploited to cheat?

Finally, I understand that work from the 8012 version will be discarded after the cutoff date, but what about work that occured before? Will it still count, or also be discarded?

Russ, NCNE

 

[ViRGE]

1) Yes. Per my understanding, if you knew enough about the bug, you could use it to cheat the system. This was allready accidentally happening anyways with some people, so not just the real cheaters are guilty.

2) The Keymaster will reissue blocks done by these clients, but you will still get credit for anytihng you turn in, in the RC5 stats up until the cutoff date; and all blocks allready done will stay in your total.

 

[Russ]

So, what that means is that if somebody did exploit the bug, they still get credit for the bogus work.

Russ, NCNE

 

[ViRGE]

Yes. Luckily though, by the time anyone figures out how to exploit the bug(as far as I know of, this is another &quot;rare case&quot; bug), Dnet should have quit accepting blocks.

 

[otlg24]

In fairness to d.net I don't think this is something people could exploit to cheat.

As far as I can tell they still have to live through ogr work times... actually I think I know how you COULD exploit it, but I'm not going to say.

Steve

 

[Adul]

oh crap, does this effect the AIX version fo this client??

 

[Poof]

oh crap, does this effect the AIX version fo this client??

Adul - the only 8012s that I had were Linux, so yeah, I expect AIX would be affected as well.
And BTW, all of the 8012s have been pulled from the download page.

 

[DJ_D]

^ bump