[plt-scheme] Re: some philosophical musings on debuggers

From: Neil W. Van Dyke (neil at neilvandyke.org)
Date: Fri Sep 17 15:01:04 EDT 2004

Some reasons that I think that development tools discussion
usually can't go beyond speculative philosophizing:

Software productivity metrics are hard,
cognitive introspection is harder,
cargo cult myths are de rigeur.

And on a related note...  Having spent much of my early industry career
developing development tools, and been acquainted with some of the names
in academe, I'll also mention a little secret: not all of the tools
research innovations one sees published have actually been built to a
functional state, much less evaluated.  Furthermore, occasionally you'll
find that a recognized expert in the field is actually personally
incapable of writing code that works (perhaps they once could, but I've
seen it devolved to what seems a clinical extreme).  Yet they continue
to successfully regurgitate old papers as new ones, cite old work that
never worked, and occasionally write up something new that doesn't work.
Sometimes I see these papers cited, and wonder whether the citing author
was misled by the papers' stated results.  A wise professor once told
me, as I was talking with her about transfering to a grad program in a
somewhat different academic field, (paraphrased) "Be sure you do solid
work, and be skeptical of what you see there."  I found that an open
mind was at least as important as skepticism, but I had no idea just how
real and systemic the causes for skepticism in some circles.

One of the nice research qualities of PLT is that code is actually
released (and often production-quality, no less) -- and therefore can be
evaluated, at least anecdotally and (with a growing user base) perhaps
as social anthropology.


Posted on the users mailing list.