[ On Mon, January 25, 1999 at 14:01:05 (-0500), Todd Vierling wrote: ]
> Subject: long options in stuff other than pax
>
> Well, I wasn't answering that particular question - and I prefer to remain
> agnostic.  (I was answering that `tar, since it has been GNU tar for ages,
> whould not break people's own tar setups frivolously.')

The problem with supporting "GNU long" options in pax's "tar" interface
is that people will think they're still using GNU Tar, and there may be
enough subtle differences (or indeed even unsupported options) in pax
that will cause them even more grief than just fixing their scripts or
whatever in the first place would (i.e. it's worse because the problems
are more subtle and may not be noticed until it's too late -- we're
talking about some fairly critical things here).

I don't have any concrete evidence yet for such differences, but I do
know that GNU Tar is a rather strange beast on the inside and that it
does have a number of unique implementation bugs that would not (could
not and should not) be duplicated by pax.  It is also true that GNU Tar
evolves, and evolving the pax equivalent of it's command-line interface
could end up being a never-ending maintenance headache (it's one thing
to maintain a local copy of some third-party code, and quite another to
mimic it).

I.e. there are a whole bunch of reasons why a full and proper transition
from GNU Tar to pax/tar would be a far better solution than hiding
behind a foreign command-line interface.

There's also the issue of GNU Cpio vs. pax/cpio, which is almost exactly
the same and tied intricately together with the pax/tar switch, with the
exception that I don't think any of the NetBSD installation tools (or
any other NetBSD tools, for that matter) use GNU Cpio, and certainly
don't use GNU long options with cpio.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>