I just wanted to add to this thread that ext3 has had its share of problems, too. I believe it is stable now, however. I've been using reiserfs for several years, on several machines, and never lost any data.
ext3 and reiserfs have notable advantage over jfs and xfs: they both support data=ordered. jfs and xfs are the equivilent of data=writeback. You can end up with files ending with a bunch of zeros, etc - writeback mode only protects metadata structures. You can lose some file data. Ordered mode will not update block pointers, bitmaps, etc until the data has been written. jfs is nice on older machines were low latency isn't an issue. jfs sucks for using xmms (so does ext3). xfs is better. xfs is more cpu intensive, and it's a much bigger module, but has delayed allocation and seems to have lower latency (no stuttering when xmms plays a flac and another reader is active). ext3 does physical block journalling, instead of just logical journalling. It typcally uses more journal space per transaction, but physical journal does allow merging when several changes are made to the same block (changes that don't have interdependency on changing other blocks). The advantage of phys journalling is that if the drive loses power during a write to a metadata block (corrupting the sector), ext3 will rewrite the whole block during recovery. Mid-write failures are more likely on some drives than others. They cause a read error until the sector is rewritten with a good crc. -- Tom Vier <[EMAIL PROTECTED]> DSA Key ID 0x15741ECE -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
