Re: FAQ questions

[Roger Burkhart]

As Nelson indicated, we've been implementing multi-level swarms as an
integral part of experiment control.  Display operations run in a top-level
swarm, independent of operations occurring in a nested model swarm.

As part of this implementation we are also finalizing, and simplifying,
our basic interfaces to Swarm creation.  So, in response to your questions
about the swarm creation code in the Jan. 96 grid/mousetraps.m example:

> 1) Are the following two groups of statement equal?
> 
>   worldSwarm = [Swarm create: aZone];
>   startupPlan = [worldSwarm getSwarmPlan];
>   [startupPlan createActionTo: worldSchedule message: M(start)];
> 
> vs.
> 
>   startupPlan = [SwarmPlan create: aZone];
>   [startupPlan createActionTo: worldSchedule message: M(start)];

There is no "Swarm Plan" or startup plan any more.  Instead you just start
schedules inside a swarm.  You can subclass a generic Swarm superclass
and provide your own method (called activateIn:) that starts whatever
is needed inside the swarm that is being started.  More details will be
in the release.

> 2) What is the relationship between "worldSwarm" and "experimentSwarm"?
>    Why both are using the "startupPlan"?

startup plan is the old way you started the things that were first running
when a swarm started.  Now you just write your own code to do that.  In
the example, both worldSwarm and experimentSwarm were using a startup
plan to create the activities that ran inside the swarm.  In the new
experiment control structure, we still have a world swarm that runs its
own simulation schedule(s), and a swarm that includes the world swarm
along with any display schedule(s).

> 3) Could you explain the meanings of the following statements?
> Are they redundent?

Those statements used messages of the old-style swarms that are no longer
available.  No, they weren't redundant, but to see how to do things in
the new-style swarm startup, you should really look at the new examples
that will be part of the release that Nelson mentioned:

> I'll let Roger answer your particular questions, but I wanted to let
> our Swarm users that the next release, the public beta on the 13th,
> will contain real live examples of multiple swarms.

I have examples of the old grid directory of test programs that have
been converted to the new swarms.  I'll try to get these included in the
upcoming release so that you can look at them in addition to the model
vs. observer swarms that Nelson mentioned as being built in to the new
experiment control structure.

Roger