Wikipedia has a nice description of journaling at http://en.wikipedia.org/wiki/Journaling_file_system
What's being added in DragonFly, however is a bit different than the traditional meaning of journaling. Traditional journaling provides in summary a "fast crash recovery" feature, and the implementation is filesystem-dependent. DragonFly, on the other hand, has a jounaling layer in the kernel, which as I understand it is placed "above" the VFS layer, which in simple terms means you can "send the data somewhere else", besides writing it to disk.
This would allow for real-time mirroring of data, even to remote disks, or to send the data to some intermediary process which could scan it for, say, audit purposes, etc.. there are many possibilities. One of the intended features is to also provide data rollback to arbitrary points in the disk "history".
I believe there is another dfly developer working on traditional journaling for UFS, but I don't know what's the status of that.