Re: [Planner Dev] One step behind, two steps ahead.

[Kurt Maute <kurt maute us>]

On Mon, 2006-06-19 at 10:07 +0200, Matteo Nastasi wrote:
> Hi, I would share my thoughts about the possible evolutions of planner 
> with you and get your feedbacks.
> 
> Two steps ahead:
>   1) I would to see full_priority_scheduling work fine in planner.

Great!  So would I!

>   2) When a 1) will be reality I wouldn't see an application
>      that it cannot be used in the real world; a simple example that
>      troubles me:
>      - 2 tasks ("A" and "B")
>      - they start at the same time
>      - they use the same resources
>      - "A" priority > "B" priority
>      - they are long 2 weeks-work
>      For one week task "A" go ahead to 50% and task "B" remains to 0%.
> 
>      After 1 week some reason imposes to change "B" priority > "A" priority;
>      what will happen?
> 
>      In my mind I suppose that the task "A" remains to 50% and task "B",
>      starting from the 2' week, begin to grow.
>      To implement my view planner needs a data model enhancement that
>      allow:
>         - correlation between "present time" and priority modifications
>         - to store each priority modification and it's time and/or made
> 	  work before it for each task.

I think in order to do this, we need to implement the ability to
baseline the project:  http://bugzilla.gnome.org/show_bug.cgi?id=140676

That's the only way we could be sure to capture a snapshot of the tasks
such that any changes, whether they're to priority or which resource is
allocated, or whatever, the history remains untouched.

> One step behind:
>   1) to develope the previously described features in an acceptable time,
>      at least for the first devel iteration, we must inactivate all kinds
>      of task's dependencies that imply retroactions (Kurt, don't hates
>      me ;), because the interaction between retroactions and 
>      full_priority_scheduling enhances algorithms complexity very much.

Could you clarify what you mean by retroactions?  You mean maybe tasks
that calculate start date from the end date like FF & SF?

-- 
Kurt Maute <kurt maute us>