On Jan 23, 2009, at 10:23 AM, Eric Wilhelm wrote:
I don't recall claiming that it was *simpler*. The single static
numeric plan is simplistic. Given any cross-platform skip issue or
optional-dependency condition, you have a situation where the plan
becomes harder for a human to get right. In that case, I want to
program the test to react to the environment rather than shipping the
wrong plan.
I want it to be hard for me to calculate. Because once I've calculated
it I can be confident that it's correct. And you're talking about a
very small percentage of tests where the calculation is hard. Frankly,
its easier and takes less time than all the time we've spend on this
issue here, IMHO.
I don't trust the computer to properly calculate the plan. I
therefore
like being able to put it in myself. It's a test! And if the computer
calculates it, it's not a test anymore.
The @things and @stuff could be hard-coded. Do you trust the computer
to calculate the length of a list?
It's not that simple, is it? If it was, it would be done by now.
Best,
David
