http://www.winntmag.com/Articles/Art...55/pg/1/1.html

http://www.pcguide.com/ref/hdd/file/ntfs/index.htm

These two URL's touch on metadata, but never on actual "data journalling"...

But here's a quote from a forum I happened to chance upon...
"New Technology File System is a partial journalling file system which logs
alterations to files prior to modifying the actual data. However, unlike a
full journalling system, it doesn't log enough information to reverse all
changes in the event of a system crash."
--Source:
http://forums.infomaticsonline.co.uk...0&thread=18705

Would it be safe to assume, then, that NTFS, regardless of version, does not
do full journalling as it may have been rumored by someone within the
c.o.l.a. newsgroup? I will be more than willing to recant this if the
opposite manifests itself...

WarpKat <nospam@nointegrity.org> writes:

So it's consistent with the rest of the OS, at least.

You need only look at the lack of bragging about "full journalling"
by micros~1 to know that there's no way in hell NTFS can do it. MS
marketing will brag about things they *might* do in the future, if
they're not bragging about NTFS full journalling, they'll *never* be
able to do it.

In article <87brnk4ck6.fsf@dps11.gnuyork.org>, Billy O'Connor wrote:
Why would you *want* "full journalling"? Here are journaling filesystems
that either only do meta-data journaling, or offer the option of full
journaling but don't do it by default (and almost no one ever turns it on):

ext3
BFS (the filesystem Beos used)
ReiserFS
XFS
JFS

--
--Tim Smith

Tim Smith <reply_in_group@mouse-potato.com> writes:

If I couldn't afford to lose any disk writes.

In article <87hdxc2v1y.fsf@dps11.gnuyork.org>, Billy O'Connor wrote:
Full journaling is not sufficient for that, as they point out in the
ReiserFS FAQ. The application could do two writes, and the system could
crash before the second is written to the journal.

If you need to be sure that a given disk write won't be lost, you have to
take steps in the application itself, such as using fsync on Linux or
whatever the equivalent things is on Windows.

Meta-data journaling, combined with fsync or similar mechanisms, gives
applications all they need to reliably write files. Going farther with
journaling either doesn't get you anything better, unless it basically turns
all writes into synchronous writes, and then your performance goes all to
hell.

--
--Tim Smith

[snips]

On Fri, 27 Feb 2004 19:10:54 +0000, Tim Smith wrote:

Simple. A buggered write as a result of, say, a system crash (power
failure, whatever) can result in corrupted metadata, corrupted data or
both.

If the data is corrupted, fixing the metadata won't help; at most that's a
way to avoid having to run "scandisk" or its equivalents after a crash.
The bogus data remains, even if the file system is intact.

In most cases, that probably doesn't make a damned bit of difference to
most people. It might, however, if the system is being used to store
sensitive data - medical records, for example - where losing an entire set
of changes is liable to be safer and/or more easily detected than having
the data there but incorrect.

Fearing a spontaneous XP reboot, Tim Smith mumbled this incantation:

I thought journalling was primarily to ensure that no partial transactions
occured.

On 2004-02-27, WarpKat <nospam@nointegrity.org> blubbered:

This sounds as though they created an awful filesystem to replace an
awfuller filesystem. It doesn't journal in the true sense since it
can't be trusted to recover fully. Instead, you get to take a
performance hit to gain an iffy possibility of maybe being OK after a
crash.

Color me surprised.

--
The three Rs of Microsoft support: Retry, Reboot, Reinstall.

On 2004-02-27, Billy O'Connor <billyoc@gnuyork.org> blubbered:
This might explain why Ewik, an authority on the Company(tm) Line(tm)
at MICROS~1, doesn't answer the question of where a journal is stored.
It's because the answer is: all over the place.

I'd bet he already knew it, too.

--
Maybe windows needs a command called "downtime" to calculate time
wasted.

In article <zcOdnegesdeFdKLdRVn-tw@comcast.com>, Linønut wrote:
Yes, but more than just journaling is required for transactions. Something
has to define which I/O requests are part of a transaction. For filesystem
metadata, the filesystem can do that. For user data, the filesystem has no
idea what data is part of a transaction.

--
--Tim Smith

On 2004-02-27 17:59:08 -0700, Kelsey Bjarnason <kelseyb@lightspeed.bc.ca> said:


Never had that problem in VMS running on a VAX. The journaling was
excellent during a major power failure. Not even a missing keystroke.
All one had to do after the power was restored was to type in
EDIT/RECOVER filename... and you were right back where your last
keystroke was entered. The key is battery backup of either the memory
planes or of the machine itself.

WarpKat wrote:

I can't believe there hasn't been any rebuttal by the MS camp regarding
this.

Oh well, I suppose NTFS not having full journaling capabilities is true.

If you crash, you crash.

On Mon, Mar 01, 2004 at 01:23:36PM -0800, WarpKat wrote:
Lol! I can see that on billboards with that fscking pedophile butterfly
guy's picture above it.

--
GNU/Linux revenues last quarter: $1 Billion.
micros~1 revenues last quarter: $4 Billion.
It's no longer a question of windows or GNU, it's a question of *Unix* or GNU.

Billy O'Connor wrote:

heh...I'd pay good money to see that happen...lol

On Tue, Mar 02, 2004 at 02:33:13PM -0800, WarpKat wrote:
Yeah, I think I'll try to mock something up now.

--
GNU/Linux revenues last quarter: $1 Billion.
micros~1 revenues last quarter: $4 Billion.
It's no longer a question of windows or GNU, it's a question of *Unix* or GNU.