very simple datastore reimplementation

[C. Scott Ananian]

On Thu, May 8, 2008 at 2:22 PM, Marco Pesenti Gritti <mpgritti at gmail.com> wrote:
>  That's true. But I also think the FUSE layer will make a big
>  difference in this regard, at the point that it *might* be enough.
>
>  I agree anyway that more transparency at the raw file system level
>  would be desiderable and that we should figure out what are exactly
>  the tradeoffs there.

FWIW, the olpcfs design exposes as much as possible of the journal
functionality at the raw filesystem level.  The xattr metadata
associated with the file (backed up with standard tar, zip, etc tools)
is *exactly* the journal metadata, so there's no loss of data when
using standard tools & techniques.  I know I've made these points
before, so I'm not going to belabor them again.
 --scott

-- 
                         ( http://cscott.net/ )