I understand that ext3 volumes never require defragging because there is some mechanism that does this automatically. Now suppose I create an ext3 partition that is accessed exclusively by a Windows machine. Does the IFS driver for Windows include a feature to keep the ext3 volume unfragmented?

I don't even think it's necessary on windows since winNT.
So I guess: no.

I agree, I would have thought the IFS Driver controls how files are placed on the filesystem, it is almost certainly the same method used in the linux kernel. So no need to defrag.

I think it is more to do with the fileystem than with Windows itself. Before getting into Linux I did a bit of reading about it and found an article, which I cannot find now, that explained the difference between Linux filesystems, and Windows.
Basically, it said ext3 will try to find somewhere on its partition to store information as a hole, where Windows filesystems don't care, and will spread one file across the partition in millions of pieces, resulting in slower load times, and probably shortening the life of the disk due to increased travel of the moving parts inside.

There are no absolutes here. If you folks search around, there are claims that ext2/3 does not fragment, and claims that it does. Some are saying that all filesystems fragment. I just read a post where someone was saying the existence of fsck is proof that ext2 fragments. Some claim that ext2 only fragments significantly when over 85% full. I read another post saying that ext2 fragments, and but it auto defragments. If that's true, it implies that the filesystem depends on an executable process to defrag, rather than the filesystem being designed to prevent fragmentation from occuring in the first place. So there's a lot of uncertainty, and no concrete information that convinces me one way or another. I'm trying to assess the risk of configuring dedicated Windows machines to use IFS/ext3 volumes, considering there isn't an fsck for Windows.

Eventually, it will have to fragment anyway, just imagine you have a file which is 30MB, and the largest available continuous free space is 26MB, the other 4MB would have to be placed elsewhere.

I suppose it's not fragmentation alone that concerns me, but the effect of fragmentation on reliability. Too much fragmentation will inevitably cause data loss on NTFS and VFAT. I'm not sure why Micro$oft would be foolish enough to allow two different files to overlap within a single block addressible by one address.

If ext3 fragments, fine. What's important to me is whether fragmentation affects performance alone, or whether it also reduces reliability. If ext3 fragmentation compromises data integrity, then I want some reassurance that the Windows IFS driver is good enough to mitigate fragmentation.

I'm searching for answers too. I'm windows user with high loaded (p2p file exchange) 170Gb NTFS partition.. Is it reasonable to change ntfs to ext2 IFS to reduce fragmentation? Have anybody compared speeds of NTFS and ext2 IFS?

bump.......

bump, what?
Is it a good idea to switch to ext2 on windows to avoid fragmentation: No, because fragmentation isn't really a problem with ntfs. And ext2 is not a journaling file system... If data loss is you concern, stick with journaling filesystems like ntfs, ext3, reiserfs...

What do you mean by that? I've seen NTFS get absurdly fragmented.. go into diskeeper after lots if data movement, and most of the blocks are red. And I've seen FAT32 volumes lose data under that kind of fragmentation. Scandisk finds lost chains and clusters, and dumps a bunch of files in the root folder containing bits of data that are no longer attached to the original file, as a result of fragmentation. Are you saying NTFS never gets lost chains and clusters as a result of severe fragmentation?
Right, ext3 was what I was referring to as an NTFS replacement.

"What do you mean by that? I've seen NTFS get absurdly fragmented.. go into diskeeper after lots if data movement, and most of the blocks are red."
I have one win2000 server that hasn't been defragged in 5 years, and a winxp pc that goes without defragging for about 3 years now. I would go take a look at the fragmentation status, if I only cared enough.
And you can't use ext3 on windows. The ext2fs driver will treat ext3 partitions like ext2.

I don't believe you can conclude from your observation that NTFS fragmentation is non-existent, or non-destructive. If your servers have seen much data manipulation over the years, an inspection of a fragmentation map would probably be solid red. So you're at risk for data loss, unless MS has somehow mitigated the lost chains and clusters.
That's good to know.. I've been using that driver on ext3 partitions, and the fact that it's been working mislead me to think the journeling is handled. I did see data corruption at one point with that arrangement, so I wonder if failure to maintain journeling can actually cause problems. IOW, would some random and incorrect journeling metadata possibly lead to an erroneous data correction?