From: thunk <gmkoller@...>
Date: 2010-03-22T04:15:08+09:00
Subject: Re: Living with a Swarm of Boids - A report from the front

Robert,

Whatever else this project is, I think it has to be a rather "extreme
design". Will you give me that?  Development and debugging has been
slow and involved early on.  Only after multiple iterations have I
been able to simplified and the roles clarified to where things are
now.  The set of "Boids" have determined the names and functions of
most of the other project parts.  There's a strong mental model to
adjust to, but some elements of the project like the need for a
"categorization" scheme did not become evident until I was thinking,
"ok I have the results from 100 fired! boids, what can I do with
them?  Early on I had developed the concept of a "result record" and
the result record could be used by subsequent passes to build reports
- I spent a week testing it back when.  What I didn't see at first was
the use of "output panel" that acts on the results of the Boids like a
big billboard passively announcing the number of various "hits" in
various subcategories.  It all makes sense, and seems totally natural
when you see it in the context of "your domain".  I really don't think
it was so obvious when I was designing the DSLs the support the
various sections of the Boids.

In he current design there is a "transmitter object", and a "receiver
Object".  The receiver object "debriefs" the Boids and send info to
the "Control Panel" and into open "Work Sheets" (probably particular
to  my domain).  The Boid author specifies the assertions he needs to
arrive at a passed/failed conclusion.  These things build up into
something pretty interesting with depth.  Each Boid, so far, is
limited to using a single helper_class - where the "real" logic and
calculations are concentrated.  The boid is a sort of "gatherer shell"
or something - but that's the whole idea - they must be kept simple so
that non-programmer types can design them.

What's exciting about this?  Hmmm.  What's exciting to me is that the
system could grow from outside experts from a network of say 1,000 to
10,000 boids "organically" without the need to change one line of code
in the framework.  What I think should be 'exciting' to
entrepreneurial types is that this can be done using outside resources
- like a forum. So the whole thing becomes something like a "knowledge
farm".  I'm not sure where the "magic" is, its all "just ruby" but I'm
watching the output of some pretty involved problem solving getting
done quickly - and I'm no where near the 1000 Boid "tipping point"
that I consider to be the design break-even point for this system
versus a strictly "conventional design".

Then, if I'm right about the run-up from 1000, to 10,000Boids this
will be getting to pretty interesting - at least to me.

Thunk

PS There is this inherent power to the web that allows a contributor
to be anybody / anyplace / anytime.  And there is this ability to
process 10,000 of these "logic capsules"/Boids or whatever in about 2
seconds.  And there is "time value" attached to knowing something "out
there" changed that effects something I'm about to do.  This all fits
together somehow, "my domain" gets close enough to justify my efforts
to me, I think there are domains where the power of this design could
be leveraged much farther - I keep thinking of medical procedures
where lives are on the line and changes are happening everyday -
around the world. A 100,000Boid system could be fired at the records
of every patient every day for about the same resources (per hospital)
as a couple hours of video game playing.  Nothing like this seems to
be happening. Why?