19 Apr
2011
19 Apr
'11
9:17 a.m.
On 17 April 2011 18:46, Alexandre Souza - Listas
<pu1bzz.listas at gmail.com> wrote:
Interesting article. It seems to have been (exceptionally-well)
machine-translated from Italian, which leads to a few confusing bits,
but it's a good essay.
One of the confusing aspects is that it suggests a new Netware
filesystem was adopted in the interval between Netware 3 and 5. It
wasn't. The NW filesystem was the only option on NW2, 3 and 4. 5
introduced new filesystems but still supported the old one - indeed I
think you had to boot off the old format, at least at first in 5.0.
The NW filesystem was interesting in several ways. One part of this
that amuses me is that apparently the entire server filesystem code
was one huge chunk of hand-coded x86 assembly. The mounted the disk,
read it, wrote it, enforced security, the whole deal. A single module,
which by the time of NW4.x was approaching a half megabyte of
impenetrable assembler.
Novell was terrified of breaking it and couldn't extend it any
further, but without doing so, it could not break the restriction of
needing a certainl fixed amount of RAM to mount a volume of given
size. More disk = more RAM.
So it wrote a whole new FS stack, renamed the old one NW Core File
Services or something and started deprecating it. Of course the new
one didn't offer anything like the performance but by 2000 or so this
did not matter so much.
The only interesting angles I know about the actual structure of the
NW FS are scant.
As disk blocks grew bigger - the great curse of FAT16 on disks of
?Oh yes,
I'd say so. ?One such fellow insisted that files on a Netware
server were stored "pre-encoded into IPX format" on the server's drives,
which is how it could serve up file data so fast.
? Dave, I may be VERY wrong (and most times I am :o)) but I remember a
dedicated novell netware server that stored the files in a different way, to
facilitate access. Now I don't remember if it had something with
interleave/phisical cluster location or something in the way the blocks were
encoded.
? I'll try googling that, This is very old info, maybe I'm completely
wrong...
? Hum, haven't found much info, beyond the hard disk had a special
formatting for netware access...See here:
http://webcache.googleusercontent.com/search?q=cache:7X_cL0uWMLAJ:blog.eukh…
? Any info on that? 512MB and the cause of the 2GB limit on FAT16 volumes -
older FSs
grew inefficient. Novell found a way to do block suballocation on
NWFS. I don't know how, but if you enabled it, it could allocate parts
of a single cluster to different files. A big gain in disk usage
efficiency resulted.
The more general thing is that NWFS was a flat filesystem. It did not
support subdirectories. All files were on a single level. You added
"namespaces" to the volume to support clients with hierarchical filing
systems - i.e., MS-DOS 2.0 onwards and Macs from some early point
after HFS replaced MFS.
What this did was use hashing to generate a unique filename from the
pathname of the file. I have no idea how it worked internally, but
something like:
/home/lproven/docs/readme.txt => "h-l-d-readme-text =>
"8-12-4-readme-txt"
This meant that it could accommodate large numbers of
identically-named files in different "directories" and that searching
and accessing them was very, very quick. Essentially it generated
unique keys for each file, so that a database-like search operation
found them, rather than recursively opening subdirectories and
scanning each one.
This only became visible when more advanced namespaces became common -
e.g. the Mac one to support Apple filename semantics, and more
amusingly, the one used for NT 3.1 and later Win9x: OS2.NAM. Since
they'd implemented DOS-like filenames for OS/2, they didn't bother
doing it again when Windows gained long-file-name support, so Windows
clients required the OS/2 namespace to be added to a volume before it
could store LFNs.
--
Liam Proven ? Info & profile: http://www.google.com/profiles/lproven
Email: lproven at cix.co.uk ? GMail/GoogleTalk/Orkut: lproven at gmail.com
Tel: +44 20-8685-0498 ? Cell: +44 7939-087884 ? Fax: + 44 870-9151419
AIM/Yahoo/Skype: liamproven ? MSN: lproven at hotmail.com ? ICQ: 73187508