updatedb, makewhatis, everything is soooo slow

[Reini Urban]

Gerrit P. Haase schrieb:
> Yes, that is the way how real filesystems are working.
> 
> If one would use NTFS to run Linux on, the fast penguinIO would also be
> much slower.

That's probably the cause by a simple, ad-hoc un-optimized design and 
then implementation of ACL's.
Metadata overhead cannot be that slow as it is.
But this is my private opinion only.

It's not the journaling overhead, not the utf-8 overhead, and it's not a 
linked list for directories, as simple ext2 or ext3 are using.
NTFS does use optimized large directory algorithms (B*Tree),
as e.g. ReiserFS does.
   http://linux-ntfs.sourceforge.net/ntfs/concepts/
Nevertheless it breaks down on > 500 files per dir as a simple 
linked-list fs.
Which for example doesn't permit ad-hoc Maildir mail servers or 
single-file based news servers (as GMANE must use).
-- 
Reini Urban
http://xarch.tu-graz.ac.at/home/rurban/