Then build results from 50 pieces!

8/400 = 4/200 = 2/100 = 1/50, so a 1/50 piece equals the 8/10 minimum

If done in parallel, fifty crunchers could finish all 50 pieces in 4 days.

If done sequentially, 50 * 4 = 200 days (6.5 months). But if the crunching has to be done sequentially, then it will take 6.5 months whether trickled or not!

Trickling is unreliable, impractical, and (since it prevents offline crunching) counterproductive. Dump the trickle!!!

Summarizing Plan C:

* Build results from 50 pieces
* Each piece is one (1) work unit
* 4 days deadline

Now every computer, slow or fast, full-time or part-time, can contribute
Now the unreliable trickling BS has gone away
Now crunching can be done offline again
Now we can crunch happily again

"If done in parallel, fifty crunchers could finish all 50 pieces in 4 days."

I think the key principle is not getting across. The second piece depends on the outcome of piece 1, piece 3 depends on piece 2 and so on... piece 300 depends on outcome of piece 299... there's no parallel processing, but maybe I'm missing something in your ABC.

Not that I care whether the trickles method is employed or not, it's just silly that if nr 9 does not get checked in before nr 10, the rules in the validator fail... that's the issue to be addressed. But, as always, if something is not statistically significant (The fail rate), it gets ignored, and for a 15 -20 - 25 and more hour task that's quite infuriating. If it does, uncheck the box.

BTW, trickling has been employed at Climate Prediction Network (CPDN) since years and years, and everybody gets their credit for each and every trickle returned. To me it was so dissatisfying to not ever be able to finish a task, of months, always some event that would make it break, that I just unticked them and not done any since years.

From monitoring, a 30*10 series finishes in about 78 - 90 days for the bulk of the tasks (See post armstrdj explaining that some seem to be hard to complete and taking lots longer, bad luck I suppose) The -ls- take appropriately longer because they are 40*10. Happen to have access to a big database of FAH2 (the hunting tool secondary purpose), which allowed me to track _1 through to _30, and if there is high _nn on the DB, go find a matching early step task and compute days between the 2 results, and extrapolate to total series time.

The WCG method just needs fixing on the server... that's all it requires, sort out the validation rules and returned trickle sequencing. CPDN is of entirely different scale, more so because of the enormous involvement to backup tasks in progress to not lose anything if 4 months into a 6 months taking model task something happens... system crash, disk flunking etc.

For all I care they stick all 300 or 400 trickle parts in a single task and just let them run, setting a trickle interval to 'at least' of x per nn hours. If that stops or there's a bad trickle, tell the client to discontinue and send the rest to the next client. Meantime give credit as [good] intervals keep being sent back... real time. This way, no annoyance, and full liberty to keep the PC on or power it off for however long... too long and 'no longer useful' would be transmitted on first client-server communication.

Anyway, based on the percent progress on the Research page, we're too deep in already to change the cart behind the horse. The flow/production redesign would just be too costly. WCG techs can fix this, and I'm convinced they can without discontinuing the trickle method.

Please consult whatever databases you have access to in order to determine the number of pissed-off crunchers who have dumped FAAH2 because of trickling.

Then determine how much faster FAAH2 would run if all those volunteers had not been driven away, and publish those results here.

You know, all they had to do was just keep running the original (and now dormant) FAAH project. They deliberately wasted lots of time and money digging themselves into this foolish hole!!

If that's still not enough to be a wake-up call to those behind the FAAH2 trickling fiasco, then let this be a powerful cautionary tale to other projects that will help them to give trickling the total and complete rejection that it so fully deserves.