Re: [hackers-il] at which point is code worthy of publication?
- You are looking for a definition of a LPU - least publishable
unit. This is a problem that every referee working for a peer-reviewed
journal or conference faces. I've done a bit of that and it's
tough. It's tough, because it is related to another LPU - least
potential utility - of published papers.
In the context of your question, it is even tougher because you are
going to make the decision for your own stuff. You are a rare bird -
most authors won't even ask the question. Rather, the question most
authors do ask is "what is the LPU in my field so that I can publish a
string of papers rather than one, thus lengthening my publication
list?" (the so-called "salami publishing"). Some prestigious
publications discourage that, but not all.
The result in the publishing world is an avalanche of write-only
papers. I suppose the result in the software world is an avalanche of
write-only projects or revisions. This should, at some point, put a
limit to the blind application of the famous "release early, release
often" rule, good as it is.
Beyond your own criteria based on your sound common sense (I would
drop "effort" from the list, leaving "usefulness" and "novelty"), I
don't have a recipe to offer. Maybe I'll just note something that is
more pronounced in code than in science: there is a clear distinction
between "features" (my code does something new) and "bugs" (useful,
but not new). There are clearly two schools of thought here. One is
represented by everybody's favourite nemesis, Microsoft, who
practically went on record saying bug fixes are not important because
people upgrade only for new features. Thus, in their definition of an
LPU "novelty" has a high weight. You might disagree.
I would guess that Googling for "least publishable unit" will yield
some fun things to read and ponder.
Oleg Goldshmidt | ogoldshmidt@...
First binary search algorithm - J. Mauchly, 1946
First correct binary search algorithm - D.H.Lehmer, 1960