Thanks for the info Sekerob; I'm glad to see that hpf2 is coming back online and that it sounds to have a nice point system while still being able to do more science. But I was responding to the proposed changes to the quorum system, which will not affect hpf2.

It's not that I am conceptually "relaxed" about the issue, it's that I believe you may be worsening the system, based on the data I can see. I just looked quickly at the last 60 results I've had validated. There were 3 out of 182 claims that were obviosly inflated outliers. In each case, the outlier's claim was discarded in favor the "normal" middle value, which means there was no practical effect.

Obviously, I don't have all the data and maybe there is some widespread problem that is completely missing the quorums I am a part of, in which case, something should be done. But otherwise, you are simply reacting to a small minority of misguided complaints at the risk of alienating legitimate crunchers. Now, it may be that the risk is low, because the tolerance will be loose enough to only affect the 3 out of 182 that I saw, so maybe it doesn't matter.

What I also noticed is that there are more moderately low almost-outliers and that averaging them into the quorum is simply to penalize everyone.

So again, please consider that you may be offering a cure worse than the symptoms.

I run a bank of laptops that generally run individually a given WU under BOINC in +_ 8 hours and recieve a claim around 50>80 ....
but I have one weak sister that cann't get a UD Agent to run; that under Boinc claims a legit 125 / WU.....it's running a AMD 485 with 148mg of ram ( or somthing close)with 3 gig avialable drive space...this machine isn't cheating it's just struggling....am I going to have to pullit off line?????

Sport

I am pleased that something is going to happen here.

By way of conducting my own experiment, I fired up two additional PC's, identical dell GX260's, one with winXP boinc, and the other with linux boinc.

I had thought that I would wait around a week for the number to stabilise as I got the xp machine online a little earlier, but at the last update, one machine has claimed for 1day and 7 hours, and the other 1day and 6 hours, so close enough already, and both with 5 results.

One machine totalled 1839 points, and the other 853.

Now I can see why a bit of a fuss has been brewing about linux being at a disadvantage from a pure points perspective.....


Cheers
Rob

I'm trying to get the word out to our Linux guys. They've stayed on other(also worthwhile) projects. A fix or patch here for Linux will be a great help!
Thanks!

I must reiterate knreed's original post:

This change to the points system does not correct the Windows/Linux discrepancy. It only evens out variation between different systems, not between different platforms. It also penalises cheaters. (Whether the inflated claims are intentional or not, they are totally outrageous. Anyone who discovers they are being penalised under the new system should examine their system very carefully, since something is very wrong. I have seen claims up to ten times the average (granted) credit. Under the new system, such claims will be marked as invalid.)

For a correction to the Linux benchmark, we must wait for the next BOINC version. It has now entered alpha testing, but the Linux release isn't ready yet. When it is released, I will retry the virtual machine comparison I tried before, and see if the benchmarks are more equitable. Other alpha testers will undoubtedly perform similar tests, and probably WCG will run a comprehensive test of their own.