[Info-vax] Request description of UFS for VMS person

[Bob Eager]

On Tue, 21 Apr 2009 23:27:27 UTC, billg999 at cs.uofs.edu (Bill Gunshannon)
wrote:

> >> Does this mean that "undelete" is not possible on Unix file systems
> >> because its logical equivalent to the entry in indexf is actually wiped
> >> out ?
> > 
> > Generally, yes. However, it will depend on implementation.
> 
> I don't think so.  It is more a matter of practicality rhan technical
> feasability.

I'm not disputing that, in a normal multitasking environment, the blocks
are going to be re-used soon, and undeletion is unlikely to be 
successful. Even in MS-DOS it was hit and miss. What I meant was that 
the complete wipeout of the inode would depend on implementation - there
are many different file systems out there in 'Unix' world.

> > Once again, it's important to realise that Unix is a rather vague term 
> > used to apply to a whole set of systems that look vaguely similar and 
> > may share some common ancestry. And, within each of those, there are 
> > multiple kinds of file systems, each of which will have its own 
> > characteristics. For example, there was a file system on UnixWare that 
> > stored all files contiguously and had to be 'squeezed' from time to 
> > time. 
> > 
> > Even UFS is problematical. There's UFS and there's UFS2. Which is not 
> > quite the same as FFS. Then there are features such as soft updates and 
> > snapshots, all of which have been added over time.
> 
> Yes, but if you know how the files are actually stored, there is no reason
> why one could not collect all the needed blocks and put them back together.
> Except, as I stated earlier, the liklihood that those blocks no longer
> hold what you wanted.

Of course. My point was more that undeletion would involve differing 
mechanisms depending on the file system in use.

> > So, saying 'Unix does this' is pretty meaningless. Nearly all of those 
> > systems can't legally be called Unix anyway!
> 
> Legality has nothing to do with technical aspects of the problem.  Saying
> that any version of BSD which can trace its roots back to the original 
> Unix code is "not really Unix" is just plain silly but, more than that,
> irrelevant.

The legality bit was just an aside. It remains that some systems are 
very different from others. And that's without the jumped-up wannabe 
Unix they call 'Linux', which is very different.

> >> With VMS and DOS, it was possible to undelete files because entries were
> >> just flagged as available and remained until used by another file.
> > 
> > Yes, although on VMS you might lose parts of the file earlier if it had 
> > extension headers, as that presents a bigger 'target'!
> 
> Are you saying that the blocks used by a file that has been deleted are
> not going to be reused?  Assuming a decent file allocation scheme I would
> expect the freed blocks to be in locations that were likely to be considered
> prime real estate for the next file that needed space.

Of course they are going to be reused. You keep flogging this bit, but I
agree with you! The whole concept of undelete, on most systems, is very 
shaky. My point was that on VMS, one might expect the information about 
the extents making up a file to disappear fairly quickly, when the file 
header is re-used, and it's worse with a file that has extension 
headers, because at least one of them is going to get re-used very 
quickly. The data blocks will also get re-used, of course - perhaps more
quickly, perhaps more slowly.
-- 
Bob Eager