Evolutionary or Revolutionary?
Posted Jun 15, 2006 13:00 UTC (Thu) by brugolsky
Parent article: Time for ext4?
Sadly, the discussion has tended to conflate a number of issues:
My principal question regarding Jeff Garzik's approach is whether putting work into an Ext4 that is largely compatible with Ext3 (w.r.t. things like chattr, EA, etc.), but is really a different filesystem is worthwhile, rather than taking the lesssons learned from Daniel Phillips's Tux2, Reiser4, ZFS, and experimental work such as ext3cow, and starting from scratch to create the next-gen default Linux fs. In other words, if extents are going to go in, what bundle of features should go with it? Can the format be changed in a way that makes adding features like in-filesystem COW snapshots or checksums possible without breaking compatibility? Will such flexibility grossly hamper performance?
to post comments)