Who maintains RPM? (2011 edition)

It seems to me that Solaris (and to a much lesser degree, Windows) has shown that the ability to rollback from a package management failure is easier done by using a filesystem with a checkpointing feature rather than by adding ACID features to the package manager.

With that in mind, RPM failure is probably better handled by btrfs and a care that user-generated and distribution-generated data are on differing filesystems (which brings us neatly to the FHS revision, the other topic on LWN at the moment, and the thought that /var/lib/rpm may not be the best location for the RPM database).