[rescue] Oracle making just a little harder to keep old machines in use

[Rich Kulawiec]

On Thu, Jul 08, 2010 at 09:18:36PM +0100, Mike Meredith wrote:
> Unless dump takes a zero amount of time to create the output, it
> _cannot_ produce a consistent view of the filesystem as it was at 00:01
> (or whenever you start it) - it cannot be done without snapshots. 

Yes, I know: I was one of the people who hacked on the code.  Keep in
mind that this was circa 1985 and the extant method of taking a snapshot
at the time was to unmount the filesystem, then dump it.  We were looking
for a way around that, and we found one good enough to use in production.
It's still good enough, provided it's used properly, and that same code
has found its way into all the BSDs (I checked) and into the various
incarnations of DEC's Unix.  Don't know about HP-UX or AIX.

It doesn't pretend to perfect: but it's pretty good.  That's why I've
used it in some rather large and demanding environments.

Snapshots most certainly have their uses -- quite a few of them.  But I
think we really do need a way to dump filesystem contents with more
granularity, more portability, and more flexible recovery options.

---Rsk