Windows Alternatives to NTFS?

Maidjeurtam asks: "I'm a multi-OS user. Although Linux is what I use the most these days (I run it on my primary P4 box and on my iBook), I also run Mac OS X and a Windows XP on other machines. Of course, those boxes are networked, but sometimes, I just prefer to plug one machine's hard disk into another. I often work with big DV files (> 4GiB) and it looks like I have no other choice than having a different filesystem on each of my boxes. Granted, Linux can read NTFS (Macs can too) and even write to NTFS partitions thanks to tools like Captive, but I don't like the idea of running Windows code on my Linux box. In fact, I don't want my data stored on a proprietary, closed filesystem. I've googled a bit and it seems there's no modern (free-as-in-speech) filesystem I can install on Windows. I'd love to have ReiserFS running on my XP box, for example. Am I condemned to stay with NTFS, or do you guys know of a Windows-compatible, open filesystem that I can use?"

Comments

ISO9660

[Anonymous Coward]

At my office we're very keen on standards compliance and basicly we wouldn't even run Windows until Windows 2000, because of the lack of certain critical standards supports (I'm not going to say the policy was that great, until 1997 we used the UCSD pSystem for everything. Have you ever seen a Pentium 166 running the pSystem? No? Exactly.) All documents are distributed in PDF and XML (was HTML) formats, email is strictly IMAP, etc. Internal programming has to be POSIX compliant and I've seen collegues dragged through the dirt for using mmap().

What we did for file systems was standardize on ISO9660 throughout, though we are considering a move to UFS so we have support for larger files. Windows supports it, though you have to hack the registry to get write access enabled, and we ended up writing a custom "disk format" tool to actually get disks initialised because, needless to say, W2K doesn't actually allow you to format disks as ISO9660 by default.

Well recommended. There are some neat features of ISO9660, like the VAX/VMS style "versions" for instance. Unfortunately, bog standard ISO9660 has crappy 32.3 style filenames (and for maximum compatability we're encouraged to just do 8.3), so it's not a perfect solution (another reason we want to switch to UFS.)

Definitely recommended though. It's a little slow, but everything will read it.

Re:File size

[DavidYaw]

And on top of that there is a ~32GB partition size limit.

That's an artificial limit imposed by Windows, to get people to switch to NTFS. Pick up a copy of PartitionMagic, it'll create those large FAT32 partitions just fine.

Reiser?

[atomic-penguin]

I'd love to have ReiserFS running on my XP box, for example.

Reiser is not really appropriate here, because you want a filesystem for "large" files. Reiser's strength and efficiency is in large numbers of small files.

Re:Why doesn't somebody write one?

[petard]

Many reasons:

1. The officially-sanctioned IFS developer kit is separate from the DDK and costs ~$1000 last time I checked. That's steep for someone who's hacking on open source in their spare time. Even if you get it, AIUI, it's very underdocumented and you'll take quite a few lumps before producing anything useful. And the license looks to my (untrained) eye as if it specifically prohibits distributing source code for the file system drivers you develop.

2. Windows filesystem drivers are hard. Developing and debugging is hairy and time consuming, and it's a good bit more work than writing an equivalent UNIX-like filesystem driver. For example, Windows expects the filesystem to handle globbing (wildcards), which is normally handled by the shell on UNIX.

3. Kernel drivers require specialized knowledge to develop and maintain. The people who have acquired this knowledge about the popular Free/Open Source drivers are, naturally, UNIX experts. They are unlikely to have the same knowledge about developing Windows kernel drivers and very unlikely to enjoy working with Windows enough to gain this knowledge in their spare time, at their own expense.

4. People who understand windows kernel driver (especially IFS) development don't, as a group, do open source. I have a few friends who do this, and they have actually mentioned that they view open source as a threat to their livelihoods and hope it goes away. One of them used the phrase "commie bullshit". I'm not joking.

5. The free kits for Windows filesystem development appear primarily targeted at academic use, and are not robust enough (or maybe simply not well-enough understood) to produce a production-quality filesystem. Some development would most likely need to go into one of the kits before it could be used to port a non-trivial filesystem with all of the expected features.

This all adds up to a serious lack of any "somebody" who can and is willing to write one, especially for free.

That said, there are a couple of free ext2/3 implementations available for various versions of windows. I've tested them, and the one that was read/write didn't seem good enough for use with any important data.

One company that I know of has developed a commercial implementation of ext2/3fs for Windows. It's not free, but for <$30 it may be inexpensive enough to be interesting.