"Can Software Schedules Be Estimated?" (via Slashdot)

[MaggieL]

This article at Slashdot points to ".a recent academic paper Large Limits to Software Estimation (ACM Software Engineering Notes, 26, no.4 2001) shows how software estimation can be interpreted in algorithmic (Kolmogorov) complexity terms. An algorithmic complexity variant of mathematical (Godel) incompleteness can then easily be interpreted as showing that all claims of purely objective estimation of project complexity, development time, and programmer productivity are incorrect. Software development is like physics: there is no objective way to know how long a program will take to develop."

I found both the main paper and the "Introduuction to Incompleteness" to be interesting reading. "Godel, Escher, Bach" and Yourdon/DeMarco fans (double points for being both) should be intrigued. The paper itself is
Mathematical Limits to Software Estimation

[russotto]

Heck, programmers have known this for a long time. We just pull a number out of the air and double it. Our manager then doubles that, his manager cuts it by 10%, the manager above him sets a deadline date completely unrelated to the estimate, and all the paperwork is fulfilled.

I've always thought the impossibility of an estimate was related to a theorem besides Goedels, though -- basically, I think the difficulty of doing an estimate is no less than the difficulty of doing the job in the first place.

added later

I just checked out the link. Turns out he does address my objection, and agrees with it, with respect to "program correctness".

Since the specification fully specifies the behavior of the program, the complexity
(C) of the specification must be at least approximately as large as the complexity of a minimal program for the task at hand.

This is something I've been saying since I learned of (and started doing) proofs of correctness as a freshman in college. Ahh, 'tis good to be vindicated. I don't suppose the Software Engineering Institute is likely to say "sorry we wasted your time" and close up shop because of this paper, though....

[Undertoad]

I have nothing to add, I just want to say that Yourden was one of the biggest Y2K doomsayers. He had previously written about the decline and fall of the American programmer in 1992, after which American programming boomed like mad. So I wouldn't exactly follow this guy into a firefight.

[MaggieL]

Originally posted by Undertoad
I have nothing to add, I just want to say that Yourden was one of the biggest Y2K doomsayers. He had previously written about the decline and fall of the American programmer in 1992, after which American programming boomed like mad. So I wouldn't exactly follow this guy into a firefight.

Well, even though some of it runs counter to the thesis of the subject paper, I like DeMarco's "Controlling Software Projects". Yourdon is a bit of a nutcase if you let him write about anything he doesn't really know about. Is there anything he knows about?

Yourdon Press publishes a lot of good books though. Not as manay as Dorset Press though.

To me, the notable thing about the paper is that it provides a formal underpinning to what "programmers have known for a long time", intuitively.