[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Switching the default /bin/sh to dash



John Goerzen wrote:

> On Wed, Jun 24, 2009 at 05:51:58PM -0500, Raphael Geissert wrote:
>> Side effects:
>> * Errors caused by the use of bashisms.
> 
> And the really important side-effect is that user scripts on all sorts
> of installed systems could experience trouble.

You are right, you made me notice I forgot to mention something:
* the sh symlink won't be modified on existing installations

> 
>> * Faster boot, builds, and general usage of /bin/sh scripts.
> 
> That could be accomplished by just setting the bangpath different,
> right?

Of every single /bin/sh script? sure

> 
>> * Reduced memory footprint when running /bin/sh scripts.
> 
> That's a nice thing.
> 
> Both of these are nice to have, of course, but we need to make sure
> this is documented explicitly, well, and in BIG UPPERCASE LETTERS in
> the release notes.  This could be a significant issue at companies
> that have done a lot of scripting outside the archive.

Of course this is going to be mentioned on the release notes (just like the
switch to rsyslog as the default syslogd); but like I just said, the change
will only immediately affect new installations.

> 
>> * It's been more than three years since Ubuntu made the switch,
>>    without all the extra bashisms-hunting Debian has done,
>>    without rolling it back.
> 
> Not necessarily relevant.  They have a reputation (whether deserved or
> not is a different discussion) of breaking people's machines on
> upgrade, and don't have the server penetration Debian does either.

It's relevant in the sense that
* we are not making a "never done before" move,
* many issues have been fixed because of their change,
* many upstreams are now aware of bashisms.

I won't object any of your other statements, but in this case we are not
discussing if what Ubuntu did was right or wrong, nor anything else. And as
such, what matters is that they made the move and it has worked for over
three years.

Thanks for your comments, and I hope that with the extra clarification I've
addressed your concerns.

Cheers,
Raphael Geissert



Reply to: