> > IIRC ext2fs doesn't have fragments, but some quite similar concept.
> 
> No.  Ext2 uses smaller blocksizes and tries to allocate them more
> contingous instead.

default block size with new linux dists (2.2 based) is 4kB instead of 1kB
so we have lot of waste. with 1kB blocks it gets VERY slow on large
volumes. especially deleting large amount of data is slow and chokes
machine.

with NetBSD you have 8192/1024 default, but 4096/512 works very good too
(at least for me)

> recovery?

because this means filesystem with files of rubbish, but in consistent
state.
 
> > fscked :)
> 
> I usually lose more data on NetBSD - but fsck doesn't even NOTICE
> all lossage.
you lose data that was not yet written. what's bad?