--- Log opened Thu Mar 15 00:00:52 2012
09:02 < mcallan> conseo: i thought the scheduler had a single job to perform, throttling the crawl so it doesn't kill the remote archive server.  this SlicedScheduler appears to do other things.  i don't understand its purpose
09:02 < mcallan> http://zelea.com/var/tmp-public/s1/votorola/a/diff/harvest/SlicedScheduler.java
09:11 < mcallan> looks like it's mostly concerned with marking.  i thought scheduling and marking would be separate and each fairly simple.  i'll wait for you to come online
09:30 < conseo> mcallan: this can be seperate, i am rather asking you about the marking process. if this is all clear besides the melting, i will separate the scheduler and return to you
09:31 < conseo> it is questionable if we really need a separate scheduler, but the marking and understanding the timelines need to be done for every forum and can be complicated
09:31 < conseo> i have tried to model it after our talks (counter + date ordering) and keep it generic enough to be reused
09:32 < conseo> but i want to share it with you (like the rest of the already documented package)
09:38 < conseo> i also want to keep it simple and as it is a specially modelled tree, i might be doing it wrong
09:39 < conseo> but mind you, i will come back when the scheduler is separate :-)
10:03 < mcallan> conseo: ok,  or if you can express the marking format in a short napkin sketch (i thought it would be very simple, like year+month+index) then we can discuss just the design - leave the impl till after
10:08 < mcallan> i would begin with a single update job, and just write down the steps it goes through: 1 read the marker; 2 ...  then you'll know what the marker needs to store
10:14 < mcallan> (it's actually not simple in the details... all the more reason to express the design before coding)
--- Log closed Fri Mar 16 00:00:08 2012