Automatic deletion of loopback device upon umount?

Rob Landley rob at landley.net
Mon Apr 16 17:39:06 UTC 2012


On 04/16/2012 07:16 AM, ralda at gmx.de wrote:
> Hi Rob !
> 
>>> Script compatibility.
>>
>> Migrating to busybox, sure.  Migrating from busybox: not our problem.
>> ...
>> Works with busybox, breaks with big distros: distros have a problem.
> 
> No! Rob, if this statement would be right, it will legalize any kind of
> modifications within Busybox as long as scripts do not break when run
> with Busybox.

Yes, it would. Your point?

You're aware there are _other_ criteria, right?  SUSv4, size, speed,
simplicity, and other things beginning with "S"?

> May be there is a very little peak of right in your words, but the
> major part is not very nice behavior against others.

The gnu project? Obi Wan Kenobi had a good quote about them, right
before entering Mos Eisley.

You're aware that Richard Stallman is the one who came up with
"POSIX_ME_HARDER", and said "The RFC forgot to send an army with you, so
it cannot expect to be obeyed", right?

  http://lwn.net/Articles/449524

Look, it's Denys's call these days, not mine.  I'm just explaining _my_
views on the subject.  Obviously he disagrees with them.

> Sure, we can close
> our eyes, but I do not feel comfortable being that blind against
> producing problems for other distro/system maintainers.

They have the option of using busybox for their command line tools if
they like.

> That is, I agree with you, we shall look for a better solution, but I
> tend to be more with Denys to try not necessarily breaking other things
> in a really bad way.

So Konqueror was wrong to introduce tabbed browsing before Mozilla had
it, and Firefox was wrong to introduce tabbed browsing before Explorer
did? Because it makes the old thing look bad?

It's certainly a point of view.

> We shall really look to get pressure to fix things
> in upstream and follow with Busybox as soon as they change.

I never considered "upstream" special, it's just something with a lot of
users. So were Solaris and AIX at one point. The GNU project didn't
invent "cat" or "rm".  The util-linux guys didn't invent mount. There's
awk nawk and gawk, _and_ a standard for it in susv4.

Let's look at a real world example of what you're complaining about:
Ubuntu switching from bash to dash was stupid. Not just because it broke
the kernel build, but because bash had been the default shell of linux
since before it was called "linux". Bash was the first program the linux
kernel _ever_ran_, in "Just For Fun" Linus Torvalds explained how he
extended his terminal program to implement the system calls needed to
run "bash" so he didn't have to reboot back into minix to fiddle with
his filesystem when he wanted to download files from the university vax
machine, and _that_ is what set him working on an operating system
kernel. Linux was literally created TO RUN BASH.

Ubuntu's reason for switch off of bash was dumb. When I objected to this
(at length), I wasn't blaming bash for adding new features, I was
blaming ubuntu for making a moronic technical decision: they did it to
speed up their init scripts (which didn't work, so they introduced
upstart).  Yes, moving #!/bin/sh from bash to dash was considered _less_
intrusive than putting #!/bin/dash at the start of each init script, and
_then_ they rewrote everything with upstart after that _failed_, but
didn't revert the #!/bin/sh symlink and admit they'd screwed up.

You seem to be implying that the GNU guys were wrong to ever implement
anything that other shells didn't, because doing so made those other
shells look bad, and inconvenienced people when Ubuntu acted stupidly.
Me, I blame ubuntu for acting stupidly. Yes, I think bash has bloated
out of control over the years, but that's a separate issue. Lots of
other shells have picked up stuff like -o pipefail long after bash
implemented it, bash didn't _wait_ for everybody else to come to an
agreement before implementing its own feature. Projects pick up features
from each other all the time.

You see busybox as a subsidiary project that only exists in the shadow
of other projects, and that stepping out of that shadow would be a
mistake. I always saw it as an independent project, and when I worked on
it I had the goal of replacing those other projects entirely, by being
better in all respects, at least by my own idiosyncratic critera.

These days, I work on toybox, and am just a user of busybox. My design
philosophy of making the best software I know how to hasn't changed, but
apparently this goal is not universally shared.

*shrug*

Rob
-- 
GNU/Linux isn't: Linux=GPLv2, GNU=GPLv3+, they can't share code.
Either it's "mere aggregation", or a license violation.  Pick one.


More information about the busybox mailing list