David B Lightstone wrote:

>(3) Mr Beck may have determined that the customer can afford
>to take the risk because the expected loss associated with
>the pruned test cases is acceptable.
>
 From my reading of XP, this is probably correct. XP Explained opens 
with an explanation of risk management and the use of an "options 
calculator". (It's been a little while since I read it.)

When you approach the problem as risk management I think you get closer 
to Beck's intent and what customers want. When customers can't define 
their needs well, or keep changing their minds, or their business is 
changing under them, we programmers need to be able to change what we 
are writing for them.

There's a certain amount of unit testing that allows us as programmers 
to more confidently change what we've already done for the customer. 
Actual practice shows that a combination of coding standards, good use 
of OO, unit testing, and acceptance testing together lower the cost of 
making changes for the customer. And some other stuff that has worked 
for the big XP proponents.

Getting too focused on any particular aspect of the process, like unit 
testing, results in a loss of focus on the big picture, delivering 
exactly what the customer wants, even though they never know exactly 
what they want before you start.

Unit testing is about delivering more quality for less money than you 
could without it. It's about knowing when you broke something. It's not 
about a 100% correctness guarantee.

The kind of testing Beck advocates is more art than science. It requires 
some gut instinct to know what to test and what to assume is unlikely to 
break. Sometimes you are wrong. You learn. You fix it. There's places 
maybe like NASA where more discipline is needed. For the small to medium 
sized business app or web thingie, that kind of discipline is overkill 
and very expensive.

IMHO. :-)

Darrin