> Whenever you see the BOINC client requesting "X" seconds of work, it is
> sending a request to the BOINC server for several work units (SETI@home/BOINC
> keeps - if possible - several work units cached in case of outages, that way
> your computer stays busy). In my experience, the computer has the ability to
> process all those work units before "X" seconds have elapsed, therefore, the
> cache is an estimate of "X" seconds (86,400 seconds = 1 day).
Close. The amount of work requested will be exceeded by up to almost one full WU. It should not quite be completed in the time requested. If you have a multi processor machine, and one CPU is out of work, but the low water mark has not been reached, then BOINC will ask for 1 second of work, and will get 1 WU no matter how long it is scheduled to take. BOINC will refuse a WU that will take longer to process than the deadline.
> Finally, you could say that the work units that your machine is processing
> using BOINC are "designed" for SETI@home/BOINC, but in reality they contain
> the same type of data as the SETI@home/Classic work units. The difference is
> that SETI@home/Classic has a "one size fits all" work unit, BOINC is supposed
> to send your computer work units that it can process in a reasonable (if not
> small) amount of time based on the computer's bechmarks.
Again, close. Different projects have different WUs, and some projects have several different WUs. If you machine is capable of crunching a particular type of WU before the deadline, then your machine isa candidate for downloading that type of WU. S@H currently has only one type of WU just like S@H1 had only one type of WU. Predictor has 2 types of WU which take very different times to process. All of my machines have downloaded all of the types of WUs for these two projects.