[WikiEN-l] How is validation supposed to work?

[stevertigo]

Without having read anything specific yet (why speak
from knowledge, when I can speak for the much larger
opposing constituency): 

The arbitrary functional values and how to process
them are not really significant as long as they remain
consistent within the timeframe that they are active.
But because changes in the processing directly and
immediately alter the nature of the program, its not a
trivial problem that a data collection test would be
drastically different from an actual implementation.

I kind of agree with Brion's criticism in this regard,
that having some data doesnt do anything useful and
therefore lacks purpose. In that regard, such a thing
in and of itself can be dangerous to even test -- if
only because it sours the concept in people's minds
without actually putting into a functional
implementation. Even if vandalism isnt the problem,
its still a problem of a proposed major functionality
being in actuality only 'free spinning wheels without
any traction.'

Its clear that there are some general ideas for how to
use the data, but what to do with the data is
something that needs to be part of the test
implementation; What's the point of taking a car to
the test track if its just a shell with suspension,
and the engine still back at the shop?

Ie. its not clear that any effort spent now on pushing
it uphill wouldnt just end up with it rolling further
back down the hill. The first point of any test run is
to test the engine -- the other stuff is secondary.
Such a 'spinning wheels' test wouldnt even be suitable
for simple -- meta maybe -- and would be expected to
fail in giving any accurate results which could be
applied to a functional model.

Stevertigo


		
__________________________________ 
Start your day with Yahoo! - Make it your home page! 
http://www.yahoo.com/r/hs