>  * Microsoft references in documentation:
I will address those in the near future. This weekend I rapidly trimmed
down parted.texi -- it's a pure manual now. The content I removed will
reappear soon in the "GNU Storage Guide".

  However, I'm not sure whether *all* of the examples will make it
through.  There are quite a lot of other HOWTOs on multi-OS stuff, so
one has to take care not to duplicate stuff.

>  * real ext3: I think a fair bit of stuff would need to be added, such
> as the ability to grow/shrink inodes. It's certainly doable, but it
> would be nice to reuse code from e2fsprogs.
IMO the most important thing is xattr support.  Apart from that I got
a lot of items on my TODO that are both less time consuming and more
important compared to the remaining ext3 stuff...

>  * it is possible that the copyright preamble is a legal requirement
> of the GPL, from part 2c. The bottom of the GPL certainly recommends
> such a preamble.
Guile, for example, which is one of GNU's central projects, doesn't do
it either, so I decided to put it where it doesn't get in the user's
way.

> some comments: - blkpg.h is
>  linux-specific, and belongs in arch/ ?
You are right, I'll move it there.

> - mbr.s is labels/dos.c
>  specific, and belongs in labels/? (mbr.s is used to generate
>  MBR_BOOT_CODE in labels/dos.c)
Does mbr.s actually work, or is it just a placeholder?  The old TODO
file mentioned it did not work, so I considered rewriting it.

>  * ped_device_get_constraint() should have the start alignment (0,
>  multiplier) and end alignment (-1, multiplier), right?
Why do we need end alignments here?  Maybe I have to say that this
function is there to avoid excessive modify-on-writes with physical
sector sizes that are a multiple of the logical sector size.

>  * I noticed that testing didn't feature in the 1.7 roadmap. If any
> one has time for it, I think a good testing infrastructure would be
> very useful. It would be nice if (almost) every bug fix got a test
> case somehow.
Too bad that the developer resources at the moment don't allow more than
basic testing.
  Apart from that last time we were discussing testing we couldn't agree
on a good testing framework.
  But this topic is still on my agenda.