At 6:18 AM -0500 12/20/98, Jerry van Dijk wrote:
>> And finally, statement C's implication about cost does
>> not even logically follow.
>If you do not repeat mistakes, you only have to fix them once.

Excuse me for stating the obvious, but if the cost of running
the process which eliminates repeated mistakes is greater than the
cost of repeatedly fixing the "same" mistake,  then the statement
is not true. Since I have stated a context within which the
statement fails to be true, then the conclusion does not logically
follow unless you prove the context does not exist (good luck!)

Processes which eliminate errors are expensive. However, they
are worth it if the cost of removing errors is expensive or the
costs of unremoved errors is high (medical equipment, flight
control, etc.).

If errors are inexpensive to remove, occur infrequently and
either are all removed or don't incur significant cost in
operation, then certain processes might not be needed.

And of course, I am not proposing the extreme: No process at
all. Instead, we should apply scientifically proven processes.
For example, several forms of peer review show extremely high
improvements in quality for relatively low cost. So this form
of process improvement is used both formally and informally
throughout the industry (and is included in CMM, although too
imprecisely to assure repeatable results):
  Software Inspection (An Industry Best Practice), Wheeler,
  Brykczynski, Meeson. 1996

>> also looking for the experimental evidence which correllates CMM
>> level with low cost and high quality. Then there will need to be
>> further research to indicate cause.
>Changing the focus from the product to the process is a first but
>important step in increasing your quality.  Before doubting this,
>you might want to study some basic literature on quality control
>and system enginering.

Ahhhhhh, well I am open to suggestions. Make sure the correlation
is between CMM level and an objective measurement of cost and
quality. The basic literature as of 1997 is very scant on details
(as is most of our e-mail) and lacked control or measurement over
most variables. Maybe I'm unaware of some recent study which has
settled the issue. I'll await everyone's more expert recommendation.

If a CMM level 5 process ever produces poor quality or increases
cost, then the assertion is false (or at least statistically less
significant). Or if a CMM level 1 process ever produces high
quality at low cost, then the assertion is false (or etc.).

And I'll reiterate an earlier point to avoid any confusion -- if you
define everything that improves quality and reduces cost as a
process improvement, then the statement is true and uninteresting.
The question is what processes (CMM characterizes a class of
processes) cause higher quality and lower cost and in what context.

>Of course, CMM, SPICE, ISO 9000, etc. are not the 'silver bullet'.
>Neither is Ada. But it sure helps.