I have a notebook (P4 2.6 GHz, 512 MB RAM), which can run computations nearly 24/7.
However, I can hardly connect it to Internet from monday morning to friday evening, because I have no direct Internet access where I live (I'm a student). Therefore, unless it is given a WU that requires days to complete, my computer has usually nothing to do starting from Tuesday, since most WUs require less than one day to complete.
Therefore, it would be great if I could try to queue work for several days (by downloading more than one WU and/or selecting WUs that take a long time). Of course, from the server side, it requires means to evaluate the time needed to complete a task - and unfortunately, file size is not the relevant criteria, since the smallest VU I got (the one I'm talking below) took by far the highest amount of time.
I know the Mersenne project, for example, allows both solutions I'm describing above. At least, the result file's naming convention ("result_0") looks like this could be possible.


On a side note, the day before yesterday, I think I lost a WU which was ~95 % done, and required ~93 hours to reach that point... I know that the completion estimation is not extremely accurate, but like all other 11 previous WUs, the WU's progress rate had been linear until then, it should therefore have taken several more hours.
tk* files in the WCG directory, and the subdirectory where the Rosetta program lies, were completely removed. I didn't check whether there was a result_0 file, but I guess there wasn't, since this WU doesn't appear in my points or # of returned results.
During the ~30 minutes before I saw that unexpected completion, the only two significant events I can remember of are:
* me saving (Windows copy & paste to another folder) tklg.ud, tkst.ud, tkop.ud~ totaling 12.4 MB.
* another program I'm beta-testing crashing.

I thought I might have made cut & paste instead of copy & paste. But checking again while another WU was running, showed that while the
files are not locked and can therefore be moved, the program survives to them being moved. Similarly, it seems the other program crashing didn't trigger any corruption (but of course, this is not
reproducable).

The corresponding WU number (ud_xxxxxx file in the WCG directory) is very near from 282916 (which is that of a task I incorrectly received, and which was therefore canceled before I received the one I think I lost - but the file remains on my HD).