On Fri, Jun 29, 2007 at 06:45:03PM +0000, Manon Metten wrote: > > It's definitely no ordinary backup or RAID. It even works with a single hd. > SFS takes care of all this. I don't have to backup anything. SFS just > writes all subsequent copies of a file to different locations on the hd and > moves the existing ones to .recycled (well, in fact it only updates the > TOC). .recycled is just a hidden directory where all previous copies of a > file are stored. > > This also means that in the rare case of a system crash when saving a > file, I only lose that part of my work that was in memory only. The copy > on disk remains untouched because only AFTER a new copy is written > to disk (to a different location), the old copy will be moved to .recycled > and the TOC will be updated. But in case of a crash during save, the > new copy isn't finished and thus the old copy remains untouched and > no TOC update is necessary. This whole process is completely hidden > for the user. .recycled only comes to mind when I have to recover some > data.
It sounds like the Log File System (LFS) that NetBSD is working on, or the database-style of a mainframe where every 'file' is really a record in a database where back copies are maintained until the space is needed. I haven't seen anything like this in Linux. There was one application where I needed this and it implemented it with a postgresql database with the files being 'huge' objects. I've never used it but you can probably use a CVS repository for this more conveniently. Doug. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
