[kplato] Summary of Data Discussion

[Brad Hards]

bilbo wrote:
> 
> On 18 Jun 2001, at 21:15, Chris Clarke wrote:
> 
> ....snip....
> 
> > b) There is an extra definition in the resource detailing units and units
> > available.  e.g. my string is defined by hours per day, and it's limited to
> > max_hours_per_day (=5 here).  For any period [startdate, enddate]:
> >
> >     extra_time = total_hours_required -
> >                           ( (enddate - startdate) * max_hours_per_day )
> > or
> >     enddate = total_hours_required / max_hours_per_day + startdate
> ObNitPick: Weekends? Is it even practicable to allow for holidays -
> probably this is a WIBNI (Wouldn't It Be Nice If.....)
I think that every resource needs to have a "availability calendar"
associated. 
People have to take breaks, test equipment has to go for calibration, and so
on.
It might make constraint fitting a lot more tricky though...

> > Depending on the constraints we're placing (of course this is complicated by
> > spreading a single resource across multiple tasks and having to take the
> > float on each task into account so that the resource is not necessarily
> > spread evenly across days, but that's beyond the scope of this discussion).
> How about having a contingency (confidence?) assignable to
> project, task, resource etc - that is every object with a time value.
> On an individual basis of course. This could be on a manhour or
> percentage basis and yet another report could compare the
> contingency of the project to the sum(?) of the components, just
> the critical path etc.
Contingency (sometimes time, sometimes dollars) is dangerously close to a
"policy" thing that we'd probably not want to code into the intrinsic
structure. It is probably better derived from more fundamental issues, like
risk. I am not sure of the exact logic yet though.

Brad