On Thu, 2004-06-24 at 03:32, Andreas Metzler wrote:
> On 2004-06-23 Ian Jackson <ian@davenant.greenend.org.uk> wrote:
> > Andreas Metzler writes ("Re: Bug#218530: 'type' and 'command -v' don't work on posh"):
> >> It does not and is not broken. type is not required by posix but is a
> >> XSI extension.
> >> http://www.opengroup.org/onlinepubs/009695399/utilities/type.html
>
> > It is however in SuS (I checked v2, but no doubt it will be in v3
> > too). Surely /bin/sh is required to be a SuS sh ?
>
> Hello,
> SUS und have been merged, IEEE Std 1003.1, to which I refered is
> *both* SUSv3 and the current POSIX standard.
>
> Of course your point is a valid one, policy just says "Thus, shell
> scripts specifying /bin/sh as interpreter should only use POSIX
> features"[1] without specifying whether we want compatibility with
> bare to the bones POSIX or something in between like XSI. Historically
> we interpreted this as "no bashism, it has to run on the other /bin/sh
> alternatives we provide, too".
>
> Things have changed. We now have a /bin/sh the implements only
> *required* POSIX features (posh) and SUSv3/POSIX is available online.
> I tend to interpret the current wording "only use POSIX features" as
> "*required* POSIX" and would suppose that XSI conformance required a
> changed wording in policy and stopping posh from being allowed to
> provide /bin/sh.
Is there actually a good reason to demand that we follow POSIX that
strictly? I can understand replacing /bin/sh with something besides bash
(which is large and slow), but dash implements considerably more than
posh, is smaller, and (IME of not running any serious benchmarks) not
considerably faster or slower. Debian maintainer scripts really won't
run right on any non-Debian systems anyway, and we don't mind e.g.
GNUisms passed to coreutils in our scripts.
We have a lot to lose by demanding POSIX-sh-only scripts (lots of work
fixing scripts, lots of work in the future writing and maintainer
scripts); what do we have to gain? I think we should set our standard at
SUSv3.
--
Joe Wreschnig <wres0003@umn.edu>
Attachment:
signature.asc
Description: This is a digitally signed message part