On Mon, May 23, 2005 at 03:22:37AM -0700, Dan Hollis wrote: > I think some of this has to do with the fact that ext3 is fundamentally a > very, very, very old filesystem. Its some of the oldest code still in the > kernel iirc. Some of the assumptions made in the original design may no > longer be so relevant over a decade later. ext2 was in use when everyone > was still on PIO and C/H/S :-) reiserfs by comparison is brand spanking > new. Not really. The fs has had major reworking over time and its one of the few very SMP scalable file systems for example. Nor is it from the world of PIO and C/H/S - it doesn't care about geometry and it threw out a lot of the old BSD cruft like interleave and rotational latency tricks that were old world. > I recall hearing something about a tailmerging patch for ext2/3 somewhere. > What happened to that? tailmerging is very 1980's, its not been worth deploying in the real world for the past ten or more years. BSD FFS did big blocks on small disks in order to get large block sizes to handle the lack of good command queueing and the latency of block requests, and then needed the tail/fragment hacks to cover up the space lossage. Nowdays nobody wants to trade 10-20% of performance for 1% more disk space given the price of disks and the disks can stream data at high speed without 8K blocks. Alan