Dale Jr, William wrote:
> That is how we do all projects - no requirements, no testing, no
> documents... ;->
> "Better get coding because there will be a lot of debugging to do!"
> William Dale
I haven't done formal XP. I have used "test-first" design on a smallish
(8K semicolons) personal project, and on implementing HW interrupt
handling for GNAT on VxWorks. The resulting unit test suites, even if
incomplete, are invaluable.
As far as pair programming goes, I only have informal experience, but
have generally found it useful.
Constant integration and regression runs are also invaluable - ACT has
had this bit implemented on a nightly basis for years. The result, in
terms of being able to respond to customer requirements, is beyond
anything I've seen in 25 years in the sw business.
So, even if not taken as gospel, many of the elements of XP have proven
to be very valuable practices, and they do appear to reinforce each
other as claimed. How they scale to larger efforts is unanswered at the
AUnit, which supports unit testing as advocated in XP, is available at
http://www.libre.act-europe.fr/ . A new release was put up just a few
Mr. Dale's comments do not accurately describe XP - but I'm assuming
they were tongue in cheek.