Didactylos,

You've missed the point entirely. What are you going to tell people 20 weeks from now when they show up in the forums and ask why they aren't getting points? Or maybe they will have figured out that the deadline changed without warning. Either way, when they tell you they would have continued crunching for nothing had they not noticed they weren't getting points... well... what are you going to say? When volunteers discover their efforts were all for naught they feel like they got dumped on and they will tell other people about it. Changing the game rules without telling all the players is really bad for PR. I'm still shaking my head in disbelief. See... if WCG will dump on them, then WCG will dump on me too. I don't like getting dumped on.

Where did I get that idea? Well, mmm, errr.....aliens abducted me and planted the thought in my brain???

Pop, fizz, glug, glug. Ok. I'm better now.

Certainly people will still continue to get all their points. I hope I didn't somehow imply anything different.

My point was simply that any result turned in after 4 days (or whatever the final reissue time is) won't really do much actual good, and will be pretty much a waste of time as far as actually helping the project goes.

People deserve to know that, and the deadline is the fundamental mechanism for telling them so. The very act of having to warn people not to cache 4 days worth of work is a major clue that the stated deadline no longer reflects reality. Out of band warnings shouldn't be needed. And most people don't read these forums anyway, so most of those who need the warning won't even get it.

Personally, I'd be much more upset if I had a slow machine and found out 20 weeks from now that my last few months of crunching had done nothing but get me points, contributing little real value to the project. But I guess most people in that situation never would find out; they'll just keep on turning in worthless week-old results and collecting points. Another case of ignorance is bliss, I guess. But it bugs me.

If all someone cares about is the points, there are other projects that "pay" a lot better than WCG, especially with the right optimized software.

Before there was never any concept of being late apart from the 3 week time out. But the quorum of 3 has been there since the beginning of the project so the slow PC effect has always been there only now that there is a realisation of a possibility of the process being more effective has this inherent effect of grid computing been presented to us. The validity of a slow PC's contribution hasn't changed by the new rules just before there wasn't a choice. Now you might get a short WU and produce a useful result within the 4 days without that Work Unit also being sent out to another quicker machine


btw the average WCG PC is just over a 2700+ Athlon churning out 3GFlops of raw number crunching power

Dave (I'll shut up now)

True, the slow PC effect has always been there, with most of the work of a slow PC being "wasted" turning in 4th results that turned out not to be needed. But at least before, a significant portion of the slow results weren't wasted; they were needed to make up for failures among the other 3 copies. So on average, 2-week old results were still fairly useful and worth turning in, in spite of the overall wastage.

But now, 2 week old results are almost totally worthless; there's practically no chance that the WU won't have been long since finished by that time. Which is great; all I'm saying is the project shouldn't still be telling people that 2 and 3 week old results are still worth crunching when they're not. It's not that there's any more wastage than before; far from it. But now we know pretty accurately which units are still being wasted, so it seems strange to let this remaining, identified loss continue needlessly when it's so easily fixed.

Before, there was a kind of decreasing percent chance that your result would turn out to be needed, the longer you took to turn it in, so the grid techs made a reasonable judgement call as to how far to let things go. But now there's a nice clean sharp line where your result changes instantly from very important to very likely useless, and I'm confused that the project hasn't recognized it. The deadlines could now be highly objective and accurate, precisely informing people where their efforts really stand, and dovetailing efficiently with the scheduling algorithms in the manager, but instead they've become misleading and counter-productive.

I'm glad the average WCG node is so powerful; that's a lot better than I expected. But there're still those who cache a lot of WUs (either always to insure against downtime, or when they go on vacation etc.), as well as those who work a lot of projects and don't give that much to WCG. I'd be really curious to see what the curve of total WU turnaround time looks like.

I'm also not so sure that none of this applies to the UD side. The only real difference there is that the points are awarded up front, and the validation happens later behind the scenes without all of us leaning over the server's shoulder and nitpicking on the method. The server still has to collect a quorum, and still has to reissue failed units, and so there's still this same kind of efficiency and wastage issue going on; we just aren't in a position watch it and whine about it. Which I'm sure is a great relief to those involved.

And so I'll shut up now too.