On Fri, 7 Aug 2020 18:02:11 -0400 rhkra...@gmail.com wrote: > On Friday, August 07, 2020 01:36:21 PM Celejar wrote: > > On Fri, 7 Aug 2020 17:21:44 +0100 > > > > Jonathan Dowland <jon+debian-u...@dow.land> wrote: > > > On Thu, Aug 06, 2020 at 12:03:56PM -0400, Celejar wrote: > > > >Ah, okay. So IIUC, each time you backup you do a full backup, and you > > > >then convert the previous backups into the reverse of the more common > > > >incremental / differential backups. > > > > > > It's an implementation detail: rdiff-backup does it too, I suspect it's > > > commonly done. From what I recall many version control systems do > > > something similar. > > > > Thanks. But what's the advantage? Isn't it more work on each backup (of > > which there are typically many), in exchange for less work on restores > > (which are typically quite rare)? What am I missing? > > Ok, I went back and did a little research. RCS, one of the early version > control systems, used reverse deltas for the main line of development -- > forward deltas for branches, it's a long story ;-). > > The intent of the reverse deltas in RCS (according to the Tichy paper) was > (is) to allow fast checkout of the most recent version.
Fair enough. But less applicable in the case of backups, since restores are quite rare, as I've been pointing out. > [[https://docs.lib.purdue.edu/cgi/viewcontent.cgi?article=1393&context=cstech] > [RCS: A System for Version Control - Purdue e-Pubs]] Celejar
