This is an interesting idea - we will have to look at that. However, this change will be sometime in coming becuase it would require a change in the client.

Buy a fishing rod. Maybe a small boat and motor too. You really need a diversion. Reordering WUs? My god man, what have they done to you?

I was thinking that the decision to stick with the "3 WUs now, a 4th later if required" practice may ultimately depend on how quickly results for the 4th WU can be returned. Having a high priority flag could ensure 4th WUs get crunched and returned with minimal turn around time. It's just an augmentation to sending 4th WUs to reliable hosts.

My idea was that the server would raise the high priority flag and that the flag would be buried and unavailable for overly zealous host operators to fiddle with in some insane effort to maximize their team's points (sheesh!)

Yes, I understand such an idea would be "sometime in coming", no problem here with that.

Speaking from the vantage of someone running a BOINC client attached to ten different projects, ... I have to say that I'd find the idea, if carried as far as permitting a project to signal that an assigned WU gets priority over other projects' WUs, is objectionable on two points:

Firstly, when I set up my own resource allocation, for the most part I give more resources to those projects that matter more to me. They all matter to me, but some are - from my own bias - worth more of my attention. Create a system which overrides the participants' judgement and I don't think that the BOINC community as a whole would respond positively.

Secondly, and probably the stronger point, the acquisition of funding and resources in world of academic and private research is already extremely competitive and - at its worst - quite vicious. Partly that's created by the scarcity of resources. Partly that's because the "small world" phenom. is perceptibly more pronounced in these circles. And several other reasons as well, but I'm not going to write a dissertaion on the sociology of science right here and now. Suffice it to say that if you piss someone off, and you'll likely regret it, as you will almost certainly meet up again at a time when they are in a position to make you uncomfortable.

Creating a system that potentially gives one project the ability to trump another in BOINC resource allocation is just begging for trouble. And I'd be a bit offended if my computer was essentially part of a battleground in a needless feud between projects that are both worthy of resources. It might not happen if a priority system were implemented, but I'd rather the temptation not even be there.

And, also, some of the worst science is borne of these kinds of feuds. Paleontology is, last I heard, still sorting out some of the misinformation created out of a rather bitter feud between two paleontologists which had turned into a contest to see who could name the most dinosaurs. I don't think the same kind of dramatically bad science would play out here, but the last thing anyone need do is give any feud any more amunition with which to fight.

(Disclaimer: All the above said might give you the impression that I think most scientists are childishly argumentative and petty. Not the case. I think that nearly as many scientists are childishly argumentative and petty proportionate to the population of us as a whole. But it's always astounding to see just what degree of damage a very small minority of particularly selfish and vindictive individuals can achieve. The only thing that keeps this fact from overwhelming the rest of us is that vast amounts of good can be done by very small minorities of particularly generous and empathetic people.)

As a postscript to what I wrote earlier - I realize I didn't write this in as I'd meant to: if a single project wants to be able to prioritize their own WUs on the fly, I see no reason why not. It would be a nifty feature. And, as you pointed out, preemption would still manage the overall allocation. (I'm a bit confused, though, on how preemption still applies if an inter-project priority system is implemented.)

I also do kind of wonder ultimately about the overall value for the effort, regardless of the geeky-coolness factor of having a priority-queue system of sorts -- why not just (as WCG has started to realize the possibility of) send priority WUs out to those machines which are fastest and most likely to return them reliably? It would accomplish much the same thing, although it wouldn't be able to get around projects with monolithic WUs (a la CPDN).

Bottom line, though, you're almost certainly right - to a degree it even takes on a tone of understatement - that it'd probably be some time before such a feature were incorporated in the BOINC client.

Speaking of CPDN, one other reason I really would not want to see one project trumping another is because of what I saw after taking on a BBC/CPDN work-unit. After about two weeks I checked the percent complete and found that it had no hope of completing before the "generous" one-year deadline. I'm not certain that the method my BOINC-client is using to figure out when it is over-comitted works for me, nor the method used for recovery. The BBC/CPDN work-unit is the only one ever in jeapordy of not being completed, and of course the one time the client detected it had accepted too much work that WU was at the end of the queue because of the first-due-first-finished reprioritization. It is only by throwing the monsterous 2000+ hour WU nearly half of the resources that I've found a way to ensure it has a chance to complete. Something else being permitted to jump ahead in line, depending on its size, could potentially put the client back in "panic" mode and potentially always stay ahead.

I'm not sure that creating a system which generates 2000+ WUs doesn't constitute some kind of arm-twisting for those of us who are attached to multiple projects and would like to allocate things at least a bit more evenly. But leaving that project switched to "no new work" for a duration after that one WU is complete should eventually balance things out. :-> If that project is still going once my efforts have balanced out closer to where I'd like to see them, then I'll take more work from them then.