Peter Seibel <firstname.lastname@example.org> wrote:
| Unfortunately it's not clear how to practice evidence-based IT--it
| seems that different software projects are more different than
| different patients a doctor might see.
I don't think that's so much the problem as that the industry as a
whole has simply stopped doing any computer *science*. That is, you
can't have "evidence-based IT" if you never collect any evidence!!
As Tom DeMarco began one of his best (IMHO) books, "You can't
control what you can't measure." [...or what you don't bother to!]
If you look back at all the papers -- and I *don't* mean from just
academia, but from industrial/commercial/governmental organizations
as well -- that were published in the 50's and into the 80's, there
were *plenty* of serious attempts to quantify the difficulty of the
task of software specification, design, development, and maintenance.
It didn't result in any "magic bullets", but then the problem was and
But the Rise of Worse is Better seems to have completely put an end
to any wide-spread attempt to closely examine the *act* of software
development in any meaningful, "scientific" way [by which I mean the
collection of predictive metrics, doing controlled experiments, even
work on cognitive models of programming, etc.].
 Tom DeMarco, "Controlling Software Projects: Management,
Measurement, and Estimatation", Prentice Hall (1982) ISBN
0917072324 [trade paperback]. Reprinted 1986, ISBN 0131717111.
Particularly interesting is his discussion of the social/political
difficulties in getting good "metrics", and his proposal for
an independent metrics team that is rewarded by the inverse
of the integral of the *absolute value* of the difference
between their estimates at any moment during the project and
the final overall actual result. [They are free to change their
estimates at any time, but the previous estimates still apply
to the interval during which they were in force.]
Rob Warnock <email@example.com>
627 26th Avenue <URL:http://rpw3.org/>
San Mateo, CA 94403 (650)572-2607