On 5/14/06, Dirk Heinrichs <[EMAIL PROTECTED]> wrote:
> What I would really like to see is something like reiser4's plugin
> scheme brought up to the VFS layer in the kernel, so that any
> filesystem could gain transparent compression.
Hmm, wouldn't that be a device mapper task, just like dm-crypt?
I don't think so, because how would the filesystem know how many
blocks are available to it? If I create a 10G filesystem on a 10G
compressing dm that averages 2:1 compression, I could store 20G of
files, but the filesystem would provide only 10G of space. Ok, so you
create a 20G filesystem instead...now what happens if you store a
bunch of already compressed files there, and run out of room on the dm
volume before the filesystem thinks you should? My guess is Bad
Things.
But if the compression is done by the filesystem, it can know exactly
how many real blocks a compressed file takes, so it can maintain it's
accounting for free blocks appropriately. Thus I think the filesystem
or VFS layer is the only sane place to implement any compression.
-Richard
--
gentoo-user@gentoo.org mailing list
