I won't rehash all of these posts but let me make two points.
- This move is not without precedent. For example the board of the journal Machine Learning (published by Kluwer) resigned en masse a few years ago to start the online Journal of Machine Learning Research. If publishers like Kluwer and Elsevier continue with their current pricing policies we will continue to see defections. Note though that Kluwer has managed to keep Machine Learning active.
- From Felten: Computer scientists are lucky, in that most of our best journals and conference proceedings are published by our professional societies at reasonable prices and terms. This is true for American conferences, most non-American theory conferences use Springer's LNCS series. For journals, the professional societies (ACM, IEEE Computer Society, SIAM) publish only a small fraction of computer science journals. While many of the best theory papers go to the Journal of the ACM, Transactions on Algorithms will be ACM's first journal devoted to papers in theoretical computer science.