Re: Linux-2.4.10 + ext3

[Andrew Morton]

Aaron Lehmann wrote:
> 
> ...
> I simply was hoping for insted of:
> 
>  <*> EXT2 fs
>  <*> EXT3 fs
> 
> (which is required today for most ext3-using people who want to do ext2
> mounts)
> 
> ... there could be:
> 
>  <*> EXT2 fs
>  <*>   EXT3 journalling extensions
> 
> AFAIK this would eliminate a lot of duplicate kernel code for ext3
> users.
> 

mm..  The filesystems could be pretty much identical on the reading
path, but they're quite dissimilar on the writing path. So the
reading-stuff code could be commoned up.

I don't think it'd buy much, though.  They are different filesystems
and the fact that ext3 borrows a lot of ext2 code is a useful
consequence of it having the same on-disk format.

And the main reason for having the same on-disk format is not, IMO, to
ease migration between the two filesystems.  That's just a once-off
activity.  The main reason for preserving compatibility is so that ext3
can leverage e2fsprogs, and the wealth of knowledge and understanding
of ext2 performance and behaviour.

The ext2-compatibility seems to be a bit of a political albatross
for ext3, really - people appear to be of the opinion that the
ext3 design was somehow compromised by the compatibility requirement.
This isn't so - ext3 is a block-level journalled filesystem.  It
could have been based on minixfs, UFS, sysvfs, etc.  Or it could
have been something altogether new.  But I can't think of any benefit
in changing the on-disk format from its current ext2ness.

-