On Fri, 14 Nov 1997 05:45:48 -0800  Jonathan Stone wrote:

> 
> Todd Vierling <tv@pobox.com> writes:
> 
> >On Fri, 14 Nov 1997 Havard.Eidnes@runit.sintef.no wrote:
> >
> >: Yes, I think it is sort of counter-intuitive that in the case of
> >: "-xpzvf" order matters, while in "xpzvf" it doesn't.
> 
> >Well, it's actually very intuitive in the case of the dash:  see getopt(3);
> >it's much more standard. In the case without the dash, you're working with
> >_historic_ tar options.  On several tars on other platforms, this is also
> >the case.
> 
> I thought we were talking about the tar we ship, not standards.
> 
> 
> The problem is that GNU tar doesn't allow mixing of long -- options,
> like unlink, and the historic usage.  And we need --unlink for
> unpacking distirbution sets.  And there's no standard (short)option
> for --unlink.  Hence, the problem for us traditionalists who've worked
> tar commands into our motor-skill sets.
> 
> You invoked standards. What do the standards have to say about the
> behavour of `tar --unlink'?  Hmmm?
> 
> Personally, I think it would be nice to fix tar so --unlink allowed
> the old-style flags. Can we do that?
> 
> If not, why not add a `U` option (equivalent to --unlink) to both the
> `historic' flags and the getopt(3) flag set?  Who *cares* if it's
> nonstandard? You just said historic tar is nonstandard anyway, so
> we're in undefined territory, just like --unlink.

Jason commited a change to tar about a month ago to _always_ use
--unlink, so if we're using a version of tar included with the
distribution tools, we don't need to worry about the --unlink.

This really doesn't help the standards argument, does it? :)

Simon.