On Wed, 9 Feb 2000 [EMAIL PROTECTED] wrote:

> This requires Ben's work to decouple the ext2 allocation size from the
> hardware page size.  We would _always_ want to write out a fragment block
> as one to ensure that the fragment descriptor wasn't at odds with the
> contents of the block.  Imagine the descriptor not being written out after
> the block was compacted.

My initial plan is to decouple the allocation size from the hardware page
size only for the cases where the allocation size is larger than the
physical block size of the disk.  Going beyond that is non-trivial, but
doable.  It may only be interesting for e2compr, since large blocks with
512 byte fragments will rock.

                -ben

Reply via email to