On May 20, 1996 at 08:27, Bill Grimsley of 3:640/305.9 wrote:

 BG> Dunno, I ran it once, just after an HPFS defrag, and it actually found
 BG> and recovered a couple of files which I'd intentionally deleted ages
 BG> ago.  At the time, I was rather impressed...

HPFS was designed from the ground up to be robust - despite what problems
people may have, it's a pretty damn clever piece of work.  Throughout OS/2
2.1 and 3.0, I've had system crashes (hardware reset style) that have never
lost a thing on any HPFS partition.  Apart from the inconvenience of not
being able to easily access 'em from DOS boot floppies, I'm now pretty damn
comfortable with the HPFS file system.

The biggest crash I've ever experienced was with OS/2 2.0 - I tried some
ray-tracing utility I'd downloaded from somewhere, and it was humming along
quite nicely .. until it crashed.  Can't remember why.  In any case, I had
to reboot (took OS/2 out with it), and was puzzled as to why the IPL
process kept failing with errors.

Boot from disks and take a look at the partition .. ooohh, yuk!  What a
*mess* that ray-tracer made of my file system!  To start with, the entire
ray-tracer's directory had replaced the contents of my "\OS2"
directory (dunno how the heck I was going to go with a ray-tracer as my
operating system..), and it just went downhill from there...

 BG> ...but didn't know it could be harmful as well.

As Doug said - in rare cases (hasn't bit me yet, but I'm in no rush to
experience, either!).

 BG>> The IBM Redbooks would be the logical choice here.
 db>> It's not there...
 BG> Oh dear, that's rather unfortunate.

As noted - IBM isn't/wasn't allowed to publish the specs so openly.

 db>> IBM is (was?) not allowed to release HPFS specs because it was
 db>> actually *Microsoft* (Gordon Letwin) who actually developed it in the
 db>> first place.
 BG>
 BG> Of course, right back in the good old days before MS decided to start
 BG> making crappy GUIs on their own.  :)

I head a radio advert this morning - something along the lines of:

    "Windows 95 calling Houston.  Errm, Houston - we have a problem;
    business just isn't interested."

    "Hang tight, Windows 95 - NT 4.0 is on it's way."

Will they ever learn?

 BG> Which is why the Gammatech Utes work so well, presumably.  It does take
 BG> quite a while when defragging to just one extent though.  :)

I figured that it just isn't worth defragging to one extent anymore; 
apparently, up to "n" extents are still stored in the fnode(?) of
the file's directory entry, and therefore occpy no more disk space - but
after that, they start using a modified B+Tree or something to manage all
the extents.

I don't recall just what "n" is, but the default for most HPFS
defraggers (3) is within that limit.

 db>> Chris Graham would be my vote for the most knowledgeable guy in
 db>> Australia when it comes to HPFS (doesn't even work for IBM *or*
 db>> Microsoft).
 BG>
 BG> Of the Graham Utilities, I presume?

The very same.  He now also has a series of HPFS tools (defragger
included), and is working on FAT equivalents at the moment (I'm on the beta
testing team - which is recruiting at the moment, BTW).

 BG> Incidentally, I was under the impression that MS's NTFS was very similar
 BG> to HPFS.  Comments?

I haven't spent much time on NTFS, but from what I can gather, they took
HPFS and added ACLs (a la HPFS386) and a journal system (a la IBM's JFS
under AIX).

Don't run Windows NT, don't care.  ;-)  (Although there's been a hint at
work that I may have to admin one or more NT boxes starting within the next
12 months - we'll see.)

Cheers..

    - dave
    d.begley{at}ieee.org

---