On Dec 8, 11:20am in "Re: shells present o", Stuart Anderson wrote: > >From the applications perspective, if only /bin/sh is guaranteed to be > present, then that is all an application can use. There is no need to > specify the other shells if they are optional. > Agreed.
> > This got discussed in Atlanta. There is a strong feeling /bin/sh shouldnt > > be posix guaranteed (since only bash 2 and the commercial ksh are). Also > > a lot of people like something small and fast running their default > > scripts. > > If /bin/sh is not POSIX.2 conforming, then somone will have to write > a full specification of it's behaviour (or at least a formal delta against > POSIX.2), as well as a test suite based on this specification. By sticking > with POSIX.2, we can leverage what already exists. > Agreed , otherwise we can only guarantee a shell named /bin/sh with unspecified behaviour. Prior to the POSIX.2 shell definition we had the XPG3 Shell spec, which is the traditional System V Bourne shell with functions. I'm not sure i can locate machine readable off hand (it was 1988) but will look into it if its of interest. regards Andrew
