On 9/4/05, David Cuthbert <[EMAIL PROTECTED]> wrote: [snip] > I've watched ReiserFS crash and burn in horribly bad ways that the > resulting fsck upon reboot took a long weekend to complete.
Was this on an ATA disc by chance? I imagine that this sort of file system would be extremely fragile with write caching. That aside, I don't see why (in theory) the filesystem should be any more susceptible to damage than similar systems without fixed-location metadata. Honestly, I think reiserfs contains some interesting ideas, and I would like to see a port someday. Though, with Matt's journalling work, the complexity of the wandering logs and atomic filesystem guarantees may be more of a detriment than an asset. After having thought about it a bit, I think the external journalling work would be a perfect fit with the WAFL file system. The main drawback with general use of this file system, is that it only produces consistency points periodically. (Which are just snapshots in themselves.) Thus the need for an external non-volatile journal. The interesting thing is that the journal is already done at the file system level, and not the block level. Its been a while since I read the WAFL papers; am I missing anything? Chris
