A new post by C. Titus Brown is worth reading: "Anecdotal science"
I'm starting to notice that a lot of bioinformatics is anecdotal.
People publish software that "works for them." But it's not clear what "works" means -- all to often either the exact parameters or the specific evaluation procedure is not provided (and yes, there's a double standard here where experimental methods are considered more important than computational methods).
This means that their result is not an example of computational science. It's an anecdote.
He gives an example and discusses the real cost, which is that a published advance really doesn't advance anything, because everyone else has to spend so much time trying to get the code to work for their projects.
Time after time I'm reminded of my conversation with the big data astronomer, who reflected that his friends who are biologists complain that students are all being trained in computer programming instead of biology. Compared to astronomy, he said, biologists don't have a data problem at all.
Clearly, bioinformatics isn't taking seriously the need to really engineer software, with documentation and standard programming interfaces.