I don't expect to see major changes to scientists' habits about publishing their software without clear incentives for them to do so (Nature 467, 753; 2010).

Given that the present value system in science is based almost exclusively on the publication of journal articles, publishing software should help scientists to publish papers. Once journals require (or at least strongly encourage) authors to submit their code as supplementary material, scientists will learn the tools and techniques necessary to get their code into a publishable state, and their employers will grant permission to publish it.

As the reproducible-research movement has also been pointing out for a while, the quality of science is then likely to improve significantly.