> > 2. PFS3 and SFS volumes *do not* invalidate. They are designed to show the
> > old disk state if interrupted while writing. However, this security
> > only covers the meta-data. If one is half-way through altering the
> > contents of a file, the bit that was being written may be wrong
> > afterwards.
>
> If PFS is interrupted during a disk write, the old file will be used
> after rebooting. The atomic update system means that the old file isn't
> deleted, nor the directory pointers changed, until the new file is
> completely written.
If a file already exists, and you alter the middle of it, PFS2 does not do
as you are suggesting. That would be terribly inefficient, and would
lead to complete and utter fragmentation, unless you are very clever
about it. The atomic update system only deals with the meta-data. Having
the meta-data safe does mean that the disk never needs checking - the
atomic-commit does not protect any more disk data than that which gets
fixed under FFS validation anyway. It just goes wrong less often.
> > 3. If you accidentally blatt various portions of the disk surface (though
> > how you would manage that is anyone's guess - I did), then you are far
> > more likely to get something back from FFS than PFS2 - I don't know
> > about SFS.
>
> PFS3 comes with recovery software.
FFS distributes the disk meta-data over the disk far more than PFS3, and
contains more redundant information. This means that if a portion of the
disk gets blatted, you can cause a lot more damage to a PFS3 partition
than you can a FFS partition. FFS recovery software has a much easier time
of it when trying to recover data.
Matthew
--
Windows doesn't just crash - it opens a window and lets you press OK first.
_____________________________________________________________
NetConnect mailing list. To unsubscribe, send an 'unsubcribe'
message to <[EMAIL PROTECTED]>
