[PRL] Code's Worst Enemy
Matthias Felleisen
matthias at ccs.neu.edu
Thu Dec 20 11:18:30 EST 2007
http://steve-yegge.blogspot.com/2007/12/codes-worst-enemy.html
I read most of it and I even scanned the comments.
The kid has his heart at the right place with his rant against the
Eclipse/Java monoculture and its high priests.
Unfortunately, it isn't more than a rant, a single piece of evidence
that something is wrong. It is an opinion and it is explicitly
written that way and doesn't even include much in terms of evidence.
I am not sure that exposing our students to such opinions -- just
because they support our own judgments -- is effective. Those who
have no experience don't understand. Those who have think they have
an opinion that's equally valid. That's what high school teaches
these days (all opinions are equal, except some (and you can guess
what I am referring to here)).
I am ccing Dan because he used to say that no project should grow to
more than N lines. We should use HO functions and macros and modules
and what have you to reduce the size. He's just as correct as Steve,
but we haven't found that silver bullet yet.
I believe that macros are one source of undiscovered power, but if
you accept macros and reasoning about programs, then we're not there
yet. If you need evidence, reasoning about naive macros (at the
source fails, fails, fails: see ACL2). We definitely need tools and
mechanisms for integrating these things better into a real language.
Ryan's thesis and dissertation with getting the specification right
and evaluating at the level of a specification are just one step in
this direction. Even if we do get this right, however, we are still
far away from proof. We must take existing projects and reduce them w/
o loss of functionality and 'reasonablness.'
So yes, Steve's rant reminds us one more time that we as teachers and
researchers are called on solving this 'size' problem.
-- Matthias
More information about the PRL
mailing list