Given the revelation that Elsevier published not just one, but six fake journals as advertising for drug companies, the El Naschie affair, the history of gun-running, and the general issue of commercial journal publishers profiting from academic research and locking away the results from the public to nobody's benefit but their own, I think it's time to take a public position that I will neither submit my own papers to Elsevier journals nor referee for them.
The two journals for which this decision would most affect me are (as Jeff already noted) CGTA, and the European Journal on Combinatorics. CGTA has been a very friendly place to publish computational geometry papers, and for now there are no non-commercial alternatives devoted to computational geometry. And EuJC is the main outlet for research in the cubical graph theory I've been working on lately. However, I'd rather have a clear conscience about how I publish my work than continue being part of the problem.
I do currently have one paper in submission to a different Elsevier journal (as well as one still to appear in CGTA); I've considered withdrawing the submitted paper over this issue, but it's already gone through a round of review and I don't want to feel like I've wasted some other academic's time over this issue. So those will stay (although I welcome feedback telling me to do otherwise) but no more.
ETA: More on Elsevier's fake-journal unit and on not publishing with Elsevier.
ETA2: John Baez does some price comparisons.