2013-06-27 I need to take some time to investigate deeply into backups. There's rdiff-backup, a seemingly good tool. I've never used it yet -- sad and bad! Unfortunately, it offers onyl the last gen- eration directly accessible, access to all the previous genera- tions needs to be done through the tool. This is not how it should be. We need the Fossil approach: read-only views on the filesystem as it had been at some time in the past. (As I've nev- er used a Mac, I don't know what Timemachine offers.) For rdiff- backups, rdiff-backup-fs [0] might provide such views. I've not yet tested it and there exists few documentation. No matter where I look, I discover FUSE everywhere, currently. This is likely caused by the too strong limitation of an hierarchical filesystem. A single hierarchy is not enough. FUSE can provide additional views on some arbitrary pool of data. That's what we want and need. This is where we approach Plan 9. While I was recherching on backups, I found Lars Wirzenius's blog posts. [1] He's the author of obnam [2], another backup software. This LWN article on the software [3] explains the in- cremental snapshots (paragraph 3), which would be used in btrfs too. Well, here we are back at Plan 9's Venti and Fossil. And the last comment (2012-06-07 by rvfh) brings FUSE back in the game. All those inventions are just old Plan 9 news. We cheer for these seemingly ground-breaking improvements, whereas Plan 9 had of- fered them already years ago when we have neglected them. How sad! Here's the relevant subset of ISO 8601 that I've waited for. [4] Markus Kuhn's comments [5] are worth a read as well. I like ISO 8601 (at least this subset) a lot, with one exception: The `T' to separate date from time. It makes the string much less read- able. An underscore would have been the much better separator be- cause it automatically suggests an optical separation. Edit: There's another thing I don't like: the 24:00:00 notation for midnight. It's like two representations for the zero in pro- gramming languages that don't use the two's complement. There's no real need for this representation, but it is an exception to the generality. If we want to have 24:00:00, why don't we have 11:60:00 as well or 11:04:60? The 24:00:00 notation violates the generality of the design, thus it should be removed. [0] http://code.google.com/p/rdiff-backup-fs/ [1] http://blog.liw.fi/posts/backups-1/ [2] http://liw.fi/obnam/ [3] http://lwn.net/Articles/500346/ [4] http://www.w3.org/TR/NOTE-datetime [5] http://www.cl.cam.ac.uk/~mgk25/iso-time.html http://marmaro.de/lue/ markus schnalke