Re: Switching /bin/sh to dash without dash essential
On Fri, Jul 24 2009, Steve Langasek wrote:
> On Thu, Jul 23, 2009 at 04:10:54PM -0500, Manoj Srivastava wrote:
>> > We want everyone to use dash by default.
>> Who is we? Why is the sysadmin not the one making the decision?
>> Why is the Vendor making this decision for the user?
> Because there's no reason for an end user to care about which shell /bin/sh
> points to. If they care, it's because they're expecting to use it for
This seems like a failure in imagination. I am an end user, and
I certainly have reasons to care. At my last job, there were production
machines that would have cared
> something beyond what Policy guarantees it to do; that's not something
> we should encourage, they should invoke the shell directly if they
> want to use other features.
Whether or not we want to encourage behaviour that does not
depend on a practice we have followed for 16 years is really
besides the point: There is an isntalled base of user scripts out there
_now_ since we have had /bin/sh pointing to bash, like forever.
For new installations, this matters slightly less, though
changing the default for a new install would make the machine behave
differently from existing Debian machines in the environment, which is
suboptimal from the user's perspective.
>> > If someone does not want to use the default, they are free to do so,
>> > but the default system shell is supposed to always be on the system.
>> Why? Is there a technical reason, or because you say so?
>> Frankly, if a user is happy with bash, they need bash anyway
>> cause they have users that use it as an interactive shell, adding dash
>> is pure bloat. They might not care for the 4 seconds it saves them on
>> boot, since they rarely boot.
> "Pure bloat"?
> $ dpkg -I d/dash/dash_0.5.5.1-2.1_i386.deb |grep Size
> Installed-Size: 216
It is still bloat. Not much of a bloat, you might argue, but
bloat it is.
You think Debian can decide when bloat is too much, or not, but
I think Debian would have a better quality of implementation were we to
let the end user decide when bloat is too much, if we can.
> On whose behalf are you splitting these hairs, exactly?
For the most important user in the world, of course: Judy, who
runs debian on her XO.
>> I think we can engineer a system where Debian suggests various
>> shells as the default shell, and the user selects one. And only the
>> selected default shell is one that can't be removed from the system.
> I think we can engineer lots of things we don't need, this being just
> one of them.
Frankly, I think that moving /bin/sh is another thing we don't
need, but thankfully the project does not run on personal opinion, nor
do we have a dictator for life making us do things one way.
> If the goal is to make *bash* removable, then I can understand why
> that would be helpful to some people since it's the heavier shell by
> None of what you're talking about in this subthread actually
> advances that goal, however. The blocker for removing bash is that
Frankly, I think you are overlooking a whole lot of things.
The frst thing you need to do is to not just make bash
removable, you need to determine of this particular user _wants_ it
too. You can't just have a limited set of scenarios (people want lean
/bin/sh) and not (people want all machines in their environment
behaving closer to each other).
> today, packages invoking /bin/bash are not required by Policy to
> depend on it. And if they did, we might find that there are Priority:
> required packages using it, which there's no policy against, making
> the exercise more or less pointless.
> Oh yeah - libpam0g is one, and libpam0g is transitively essential.
Again the tunnel vision on packages -- there are users with
installed bases too, which every one seems to just forget.
The idea I am espousing is that we need to come up with not just
replace bash with dash, we need to ask the user if they want to change
the default shell, and whether the new default shell should be dash.
Given its constituency, the only thing I expect to be "open" about [the
Open Software Foundation] is its mouth. -- John Gilmore
Manoj Srivastava <email@example.com> <http://www.debian.org/~srivasta/>
1024D/BF24424C print 4966 F272 D093 B493 410B 924B 21BA DABB BF24 424C