> You could at least apologise.

I'm sorry for your data, but I won't apologize for me working on GNU
Parted, because I'm doing my best.

> Well, I got a segfault.

At what stage of the process? Right at the beginning or in the middle?

> The version I'm using was downloaded from http://www.sysresccd.org/ on
> Saturday. So, is this a new bug?

Probably (not). I can't say. We have made big steps since the last
release (1.6.25.1 in december), and the mentioned rescue CD (as most
others) is still shipping the 1.6.23 from half a year ago.

> >parted-devel@lists.alioth.debian.org
>
> According to the parted web pages, this is a closed list. Do I have
> permission to subscribe to it?

"Closed" just means subscriber-only. Messages from non-subscribed people
get into the admin queue first.

> The big problem with bug-parted, from what I saw, was that no-one
> actually answers anything.

I know, everyone's busy at the moment and I'm the only one answering
people's issues. Also, I guess a lot of people just don't post because
they think asking for help in some non-Parted-specific way is not
appropriate.

> >I am working on the GNU Storage Guide, which also will contain
> >comprehensive information on data recovery.
>
> That's an extremely good idea, thank you very much. I've been saying
> for years that lack of documentation is Linux's problem, not a lack of
> software. Linux would be far better off with more documentation and
> *less* software.

You don't happen to have some time and enthusiasm to consider
helping...? :)

> It's only when the kernel tries to mount root that it fails.

If it's ext2/3 I suggest trying the backup superblocks.

> What I really need is something that will search for the remnants of
> directories and inodes, and piece together what it can of the file
> structure. I thought e2retrieve would do exactly that, but it seems to
> throw an exception before I get a single file back.

Try e2fsck (backup first if at all possible!), if it's ext2/3.

> >First, you don't know what the cause for the problem is. Second, at
> >several locations (documentation, in some cases the program itself)
> >Parted advises you to backup your data before working on the media.
>
> Part of my point is that, in real life, *no-one* backs up their data.

I do, and I rather have half or two thirds the space while keeping a
fallback.

> Hard disks are expensive. Yes, really. Backups are extremely slow,

Don't you sleep at night? :) Also, DVD-RW writers are getting pretty
cheap.

> especially if your best option, as in my case, is USB 1.

That's a bad option, I concur.

> But the kicker is: it is really, really complicated to set up backups
> under Linux.

I use 'rsync'. Finding a proper GUI backup tool is probably not that
easy, but I don't know much about backup stuff.

> out there, each of which requires lots of technical knowledge to
> operate

I suggest a relaxed after-noon with the rsync manual page and a set
outcome of exactly one command line.

> and none of which provide the kind of facilities a normal person would
> expect (like being able to browse for just that one file that you
> lost).

With my rsync backups I am just able to copy the lost files over.

> Assuming that your users have backed up their data is really not a
> responsible policy to take here.

No, but it's the best of the practical ones. We _cannot_ rule out bugs
(and neither can Partition Magic), and this fact together with the
sensitivity of the operative area must have the conclusion of back-ups.
Better sorry before than sorry after.

> Parted is now distributed on a huge number of live CDs that are
> pitched *directly* at inexperienced desktop users who want to play
> with Linux. These people are not reading your website, and they're
> used to ignoring warnings about backing up their data. It is far, far
> too easy to use parted.

In my opinion, people must assume responsibility for their actions, and
if they don't do it, life will teach them. Sorry for saying this, but I
don't like "restricting people for their own good". But let's leave it
at that, this is worth a whole discussion for itself, and we can discuss
more important things here IMHO.

> >Third, the license explicitly says that you are using it on your own
> >risk, as it is usual for free software. If you want professional
> >consulting or safety guarantees, you pay for it. This is not
> >different from commercial software -- the latter one lets you pay
> >doubly, though.
>
> It is very different. If I had bought and used partition magic, and it
> wrecked my hard drive due to bugs or bad design, I'd tell the world
> about it, for example on Amazon review pages. This would damage their
> business, especially if it happened to lots of people.

It damages my feeling for responsibiliy and my compassion if someones
tells me software I work on went bonkers and destroyed data, so I'm
doing my best to avoid this. Quite an incentive, if you ask me. Money
means less motivation for me. OTOH, do the many complaints of users of
Microsoft Windows damage their business considerably? Do they take pains
in making their software much more reliable?

  Perhaps you can comment my statement about paid support and
guarantees? I mean, instead of buying PM you could just use GNU Parted
and pay someone to give guarantees, backup space or advice.

> So they have an incentive to make sure that the consequences of using
> partition magic are never too severe.

So, do they succeed in this? I don't know, but I guess it did a lot of
damage for some people. Another "big player", the Norton Utilities, is
quite crappy, too.

> What's your incentive? You like programming. You like adding fun new
> features. You don't have any incentive to harden the software except
> avoiding flames - but they're easy to avoid.

That's a blanket statement. I can't remember any place where I have
stated my motivation for programming on the web or elsewhere, and I
haven't seen other people working on GNU Parted doing anything like
that. The incentive you have in mind is probably true for people working
on games, and their average software quality is evident.

  My motivation is making GNU Parted a both user-friendly and VERY
reliable tool. Features come after those two prime objectives.

> This is why, if you can suggest any commercial software that can trace
> through ext2 filesystem structures, I'd rather use that than any free
> software at this point.

I must emphasize on this: commercial software does not (neither
automatically nor usually) mean more reliable software or any other
advantage.

> >I conclude that Parted is not more dangerous than other software,
> >it's just that is touches an area of work that is likely to make the
> >user VERY upset when something goes wrong.
>
> Um, your logic seems to be that "since, unlike other software, parted
> gives no guarantees at all, it's not more dangerous".

That's not what I intended to say, and I agree with you that the first
sentence is probably not correct without some additional comments. I
had other mission-critical packages in mind -- think about the lots
of people that have lost their data due to some kernel bug in Linux,
Microsoft Windows or any other operating system. And this is just an
example.

> Actually, that would indicate to me that it *is* more dangerous.

It's more dangerous than your average text-processing application and
most other software, yes.

> I think parted probably works OK if you're a professional software
> house who as a matter of policy always backs up everything. If you
> work at SGI and HP, an instruction like "first, backup your entire
> hard disk to *another* hard disk" is no big deal. I'll bet they're
> using fresh 200GB drives as coasters. I, however, as a private user,
> don't have that luxury.

Even if they have backups, they are of course interested in the
evolution of the product that was the cause for the problem -- backups
are also expensive to restore and result in down-time. And we haven't
received a lot of bug reports from them lately...

> No, I'm saying that developers of partitioning programs *who don't
> have the resources to do a thorough job* should move to writing
> software in an area that isn't so sensitive.

Linus Torvald's and his team back in the beginning 90s would never
have written their UNIX kernel that is so widely used now if they had
followed your advice. Yes, I know you are talking about partitioning,
but in terms of danger it's the same league, maybe even more.

> As one suggestion, there is a chronic lack of backup systems that are
> actually suitable for normal, home computer users. A lack that has
> bitten me quite hard in this case.

While a lot of programmers would happily try to save all problems of the
world, they have limited time and energy :/

> If you have enough people to really give parted the attention it
> deserves, then fine, it's a great addition to the stable. But for
> something so delicate, I believe that you really have to be sure
> about it. You have to double-check and triple-check everything. If
> my instant messenger program crashes half-way through something,
> it's annoying. If parted crashes half-way through something, it's
> a catastrophe. You need to make damn sure that the thing just does
> not crash, ever. From your FAQ, it sounds like you don't have enough
> resources to make that happen.

The sentence should just state that new feature additions (like the
block size issues) are slow to come at the moment. Whenever we find
(or are reported, which is the usual way because storage issues are
very complicated, full of traps and incompatibilities) a really serious
(destructive) bug, it will be fixed ASAP and the release won't wait long
(another difference to commercial products). As a note here, the core
dump of the "rescue" command is not a serious problem, it's just that it
does not work and will abort the program. It's a non-destructive issue.

  Also, the sentence is just a bit of a rant from me because I'm at the
moment the only one giving GNU Parted significant attention.

> Sorry to rant, but like I said, I think this is an important point.

I consider this discussion to be very important, and I'm glad you are
bringing up some interesting points. If you do not object I maybe would
like to publish this educating conversation.

  Cheers,

    Leslie

--=20
gpg --keyserver pgp.mit.edu --recv-keys 0x52D70289
http://nic-nac-project.de/~skypher/

--lrZ03NoBR/3+SXJZ
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)

iD8DBQFD50MNn/ep3VLXAokRAoqtAJ0XNQXNg6YFvZOy+vtg++3fpkcZTQCg2UN2
q0qdTa1h/Xcfwdq5fudsUBM=
=n8BZ
-----END PGP SIGNATURE-----

--lrZ03NoBR/3+SXJZ--