Re: making nuthin'...and calibration tests...

[Chris G. Langton]

> :)  And such a powerful tool to increase our ability!  Just think how much
> easier and faster you've made it to make truly toe-curling ones!

Ahhh!  But as you suggest, If we couldn't make toe-curling mistakes
we couldn't make toe-curling discoveries! 

It's those toe-curling discoveries I live for....

Actually, my toes are permanently curled, so don't listen to me....

Seriously, though, this all brings up a point that I've been wanting to
raise with respect to Swarm, or to running any experiments on computers
for that matter, which is that whenever we provide a tool, we need to 
provide a "calibration" test or set of test cases that we can run 
that tool on to convince ourselves that the tool is operating correctly..

Now I know that it is impossible to absolutely verify computer code
in the general case, but this is also true for verification of
any physical process in general, including the experimental
apparatus found in "real" laboratories. Nonetheless, an important
part of setting up experimental apparatus in a "real" lab is
to calibrate and shake down the equipment via standard reference
tests etc... 

So, I'd like to encourage anyone who develops a tool or useful 
experiment object under swarm to also provide specific 
calibration/verification test cases, as well as suggestions about
the proper domain of applicability for the tool (e.g., some
random number generators are fine if you use the whole 32 bits,
but miserable if you only use the low-order bit).

Especially since software that runs fine on one platform may have
weird behaviors on other platforms, samples of "correct" input/output
mappings for the software are critical. This simply corresponds to
calibrating a scale using standard weights, or to calibrating
a spectroscope using known reference flouresceing gas-tubes.

We'll try to do that ourselves for the tools we provide....
and again, we know this can't be perfect, but it can certainly help 
head off some potential toe-curlers of the wrong sort.....

A high priority for us in the near future is to provide debugging
tools such as activity traces (what actually happened), as well
as schedule browsers and tools to examine the linkage structure
among created objects.

Chris Langton