I assume it varies by institution and field, but with the decline of centralized IT in universities, I'm seeing more open-source in academia, because the infras

I assume it varies by institution and field, but with the decline of centralized IT in universities, I'm seeing more open-source in academia, because the infrastructure and budget to buy and administer site licenses for commercial tools no longer exists. Even a mere 10 years ago, everything would've been done on either the department Unix server, or a centrally managed Windows desktop install, and both of those came with the whole range of Maple, Matlab, Sicstus, etc. licenses. But nowadays I have a Debian server in the server room I administer myself, and a laptop I also administer myself. I could buy a Matlab license for my server, but it's basically up to me, so I probably won't. The other shift is that open-source tools seem to have been slowly been chipping away at the incumbents in statistical software. People used to use SAS and SPSS and that kind of thing, or commercial AI software from the likes of Salford and Symbolics, but now everyone I know uses R, stuff from MLOSS [1], and Weka. This could be local to CS/AI, though; I do get the impression that engineering departments love Matlab. [1] http://jmlr.org/mloss/ http://jmlr.org/mloss/