> skip all the gory details, their primary benefit to ReiserFS is their astounding speed. For example, searching a tree with trillions of items takes only a few dozen operations. For ReiserFS, storing metadata in the tree meant there was no limit to the number of files a directory could contain—unlike Ext2, which got bogged down for every file added.

But this had a price: high cpu usage. When your filesystem is competing for cpu cycles, you have a problem. In practice, ext2 was the fastest for normal use. Other filesystems added journaling which made them slower on a normal system (update data + journal). Of course, as CPU speed evolved, journaling was becoming faster.

Reiserfs on a 486 was unusable.

This is an odd article about ReiserFS and has little to do with any generalizations about Linux filesystems. Yes of course Reiser failed when there was one developer and that one developer was arrested for murder. As for B-Tree XFS works just fine for handling large numbers of files/directories. There is something off about this.