writing to windoze partition

[Craig White]

On Tue, 2005-08-09 at 10:40 -0500, Mike McCarty wrote:
> Craig White wrote:
> > On Mon, 2005-08-08 at 16:15 -0500, Mike McCarty wrote:
> > 
> >>Paul Howarth wrote:
> >>
> >>>On Sun, 2005-08-07 at 16:00 +0200, Liloulinx wrote:
> >>>
> >>>
> >>>The Linux NTFS module has very limited write support.
> >>>
> >>>http://linux-ntfs.sourceforge.net/rpm/rel26.html
> >>>
> >>>Paul.
> >>
> >>
> >>IMO, Linux has limited *read* support. I have tried with
> >>two different versions of Linux to read NTFS without
> >>success. It appears to copy data correctly, but as it turns
> >>out this is not the case in my experience.
> > 
> > ----
> > of course you worked with the developers of the project to help iron out
> > the wrinkles on it before you thumbed it down on this list right?
> 
> Sarcasm is unbecoming. I did the copies on Saturday of last
> weekend and the weekend before that for a friend at her house.
> I don't normally have access to the machine unless I go over
> there. So I've had all of two days now to get everything together. 
> Furthermore, the versions of Linx I tried were not Fedora Core, so
> posting here would hardly be the correct forum. Or would it?
> 
> And I didn't write some sort of scathing criticism, either. I stated my
> experience.
> 
> For those with more interest, I found two things occurred, and
> have not investigated enough to check for others.
> 
> File names changed in case. I mean that a file stored in the NTFS
> partition with a name like FRED.EXE might be copied to a name like
> FreD.exe. In one case, the same file got copied with different
> destination file names in two copies. By that I mean that I did
> a copy from NTFS to a FAT32, and then repeated the same copy
> (to a different destination) and the same file wound up with two
> different munged names. Like FreD.exe in one copy, and FReD.Exe
> in the other copy. To clarify further...
> 
> Using a Live CD based on Debian
> 
> $ mount /mnt/hda1    (FAT32)
> $ mount /mnt/hdb1    (NTFS)
> $ cd /mnt/hda1
> $ md oldd
> $ cd oldd
> $ cp -pr /mnt/hdb1/* .
> .......
> $ diff -r /mnt/hdb1 . > diff.out
> <6000+ files had different names>
> 
> Using a different Live CD, also based on Debian
> 
> $ cd ..
> $ md newd
> $ cd newd
> $ cp -pr /mnt/hdb1/* .
> $ diff -r /mnt/hdb1 . > diff.out
> <6000+ files had different names>
> 
> Some of the files in oldd had a different name from
> the files in newd. I don't know whether using the
> different versions of the Live CD contributed to this,
> or whether repeating the copy would have had similar
> results using the same CD again. Both of the CDs were
> the latest releases of the respective packages. Names
> withheld intentionally. I'm not trying to criticize those
> packages, especially behind their backs.
> 
> Some of the file names were corrupt by having unusual characters
> in them. Like 0xC1 and 0xE1. I did diffs of the files, and in the
> files where the names were not different, the data were correct.
> But over 6,000 files (about half, I guess) had munged names.
> 
> When I have a little more spare time, I intend to go over there
> and write a script to rename the files with changed names, and
> do the diffs again and see whether the data are all correct.
> 
> In any case, I don't believe that I am obligated to help solve
> the problems, as you seem to suggest. Had I posted a criticism
> like "DON'T USE LINUX! IT SUX! IT RUINED MY DATA! IT'S A DOG!"
> you'd be justified in replying "Then don't use Linux, and please
> go away." I can think of no theory to which I subscribe which
> obligates me either to report or to help solve problems.
> Presumably you are attempting to shame me into feeling guilty
> about not being more civic-minded in some sort of fashion.
> 
> Well, if that was your goal, it didn't work.
> 
> What I did was pass along what I thought might be some useful
> information for a prospective user. Not write a criticism.
> 
> Happier?
----
happier? no. interesting that you criticized my sarcasm before resorting
to your own.

Your problems with copying the files would appear to have nothing to do
with the ntfs kernel module - but rather related to the differences
between NTFS and VFAT file name handling methodologies and Linux was
indifferent to it.

I think you would have a much easier time of things if you shared the
NTFS partition with samba to a Windows computer and allowed the Windows
computer to copy the files as it understands the nuance of the
filesystems. Granted a local copy is quicker than a network copy. But a
local copy that doesn't work vs. a networked copy that does work seems
to have greater efficacy than writing script to rename files that still
may have unanticipated results.

Craig