Re: [E2FSPROGS, RFC] New mke2fs types parsing

[Theodore Tso <tytso@MIT.EDU>]

On Thu, Feb 21, 2008 at 01:52:21AM -0700, Andreas Dilger wrote:
> On Feb 20, 2008  17:20 -0500, Theodore Ts'o wrote:
> > There are only three things which mke2fs will do, in my design:
> 
> This should all go into the mke2fs man page...

It will be documented; as I said, this was a request for comments
about the overall design, before I finish polishing it and adding man
page documentation, etc.  The patch was very much an interim patch,
including lots and lots of debugging printf's.  :-)

> 
> > [fs_types]
> > 	ext3 = {
> > 		features = has_journal
> > 	}
> > 	ext4 = {
> > 		features = extents,flex_bg
> > 		inode_size = 256
> > 	}
> 
> Presumably the ext4 feature should also have features = has_journal?
> If this is the default for ext4, why would it need to be given for ext3?
> 
> We should also add "dir_nlink,flexbg" while we are in there.

Yes, of course.  This was an example, not what I plan to check in.
(And it's flex_bg, not flexbg; we also need to add the uninit_groups
flags, etc.)

The other thing which I've been considering is some what to make the
feature list displayed by dumpe2fs a bit easier to understand.  One
thought is to bundle a number of features into things like std_ext2,
std_ext3, std_ext4, etc., with an option to display the full set for
someone who wants a more verbose/explicit description.  The one caveat
here is that once a bundle is defined, we don't ever want to change it
so that when someone e-mail's a dumpe2fs output as part of a bug
report, there is no question about what a feature bundle means; it
can't be e2fsprogs version dependent.

     		       	      	       	   	- Ted