Pixel wrote:
> here is a sort of comparison i wrote a few days ago:
>
> - resizing ext2: i removed it from diskdrake as it was buggy and ext2resize
> doesn't handle sparse superblock (yet). parted has the option but it never works
> (cuz of sparse superblock)
I didn't write the ext2 code (Lennert did), but I believe it does
support
sparse superblocks.
> - resizing raw partitions: parted do not want to resize what it doesn't know
> about. diskdrake warn the data will be lost and goes on :)
For things that look safe to ignore, it gives you the option to ignore
errors.
> - resizing fat partitions: c2perl doesn't really exist and isn't magical! A lot
> of changes were made when porting to
OK (I thought there was one! It would be feasible to write one, so why
not?)
> - geometry handling (dos compatibility): parted try very hard to get the good
> geometry. diskdrake doesn't care much about it (it takes what linux gives)
OK.
> - partition table labels: i don't really know how this work, but parted has it
> and diskdrake doesn't
How can you resize file systems without changing the partition table?
> - move partition: diskdrake's code is buggy and not enabled. parted seems to
> have only ``copy'' partition. i have not tested yet
You can do "move" then "rm".
> - ease of use: parted 0/20 (lower than fdisk), diskdrake 19/20 :)
parted fits on a rescue disk. It's light-weight. libparted is designed
to allow other front ends to be written easily. So each distro can
write
a front end for their installer.
> - warnings: none in parted, even worse, no `Write' action (as in fdisk), all is
> done after each action. One before each dangerous (lost of data) action in
> diskdrake
Are you suggesting that parted should queue operations? I guess it
could,
but I imagine more users would screw up...
Perhaps I could do an very-interactive mode, which prompts for
everything.
The thing is, even a "rm" isn't that dangerous - use gpart to recover
it.
In fact, Parted could very easily recover deleted partitions.
> - unity: parted->cylinder, diskdrake->MiB
The command-line version is supposed to be low level. It makes sense to
work with what you align to.
> - undo: diskdrake (not for all actions of course)
> - raid handling: diskdrake
I'll add this to TODO. Thanks :-)
> - fstab: diskdrake can modify fstab, parted doesn't as it doesn't ask mount
> points
Parted is meant to be low-level. Other front ends can (and will) do
this.
> - 1024 cylinder boundary detection: diskdrake warns the user if / or /boot is
> created or assigned beyond the 1024 cylinder. parted doesn't as it doesn't ask
> mount points
Ditto from above.
> - save partition table in file: diskdrake only
> - save partition table on raw floppy (at install): diskdrake only
> - different level of expertise: diskdrake
> - change type of partition: strange in parted (use mkfs)
> - extended partition handling: limited as in fdisk (ask me for more :)
More please :-) I'm very scared of extended partitions, because
different
OS's and utilities treat them differently. I know there are other
things
DOS can handle, but they're not very safe.
> - partition types: 3 in parted (not even FAT32!), many in diskdrake (and fdisk)
Parted doesn't deal with partitions and file systems separately,
because that really complicates things for the user. Therefore, FAT16
and
FAT32 are both FAT. It doesn't show the user this kind of thing.
(Perhaps
it should, I don't know... no-one's complained)
Why would a user want to change a partition from ext2 to fat,
without formatting the partition?
Parted can even convert FAT16<->FAT32.
> please correct if i'm wrong. i'm just hoping Andrew will remove that nasty
> description from parted's README.
I guess the description is nasty for the wrong reasons. I wasn't
impressed
by how your project completely ignored mine. Even after I emailed you
about
bugs in my fsresize code that you used (that I found were still in your
code), you ignored me. I haven't checked if they still remain.
You've convinced me I should let users decide, rather than be
judgemental. What do you want my description to be?
My comparison with Disk Drake is:
* Easier to use, harder to make mistakes
* More complete solution (handles /etc/fstab, etc.)
* Less support for FAT (can't resize FATs, can't convert FAT16<->FAT32,
can't copy) Perhaps you changed this?
* Less support for ext2
* Less attention to compatibility with (older) DOS/Windows systems
Are you OK with this?
I still think we can cooperate somewhat, although it is difficult after
a code fork:
* share info about undocumented things around partition tables
and FAT16/FAT32.
* share TODO items.
* discuss algorithms. I've been designing algorithms to change
cluster sizes for FAT. It certainly isn't trivial.
Andrew Clausen
