Faster filesystems is another virtue.
How much faster are those filesystems than NTFS for "normal" use, though? It's pretty hard to come up with a benchmark, considering that only ext2 is available for NT and that the NTFS support in linux can not
be compared to native windows speed (neither the current ntfs3g nor the old wrapper around binary NT drivers). Yes, I know that some of the filesystems are theoretically better, but outside insane practices like putting 20k files in a single folder, can you feel
the difference? (explorer.exe is dog slow with large folders, but that's because of explorer and icon extraction, not
because of NTFS).
Also, most of the newer filesystems on linux (those actually offering performance benefits) can't yet be considered as mature as NTFS... there's plenty of ReiserFS horror stories around. EXT2 might or might not be faster, but it doesn't have journaling - EXT3 w/journaling should be stable, but I haven't tested it's speed... and really, for normal desktop/workstation use, I don't think I've ever bumped into a filesystem bottleneck.
One of the few reasons where I can see the filesystem make a big difference (apart from data security differences or managing features, like über-cool ZFS has) would be mail servers storing email in Maildir format (which is superior to mbox)... but for most other stuff I don't believe in having a zillion folders or files inside a folder
. That said, I wouldn't mind more filesystems being available for Windows; NTFS is decent, but I'd like the data-safety features of ZFS, and I wouldn't mind seeing proper
head-to-head benchmarks, even if synthetics don't have much to do with desktop/workstation patterns.
You mean Windows 9x where you had to reinstall every few months just because things got too slow? Now who wouldn't love that!
Never really happened to me - when I needed a reinstall it was because I hosed my system, which was ever so easier on 9x than NT because of the lack of protection (user permissions as well as memory - having all DLLs in shared writable memory is bad). Back then it did
mean a measurable performance difference to clean & compact your registry, though