Re: .rpm? .lsb??

["H. Peter Anvin" <hpa@transmeta.com>]

Alan Cox wrote:
> 
> > Obviously, the LSB should specify a particular version of the package
> > format, and quite likely of the rpm command (if it is included).  This
> > is no different than specifying a kernel version, C library version,
> > etc.  Alternatatively, the LSB can specify an interface which happens
> > to correspond to a particular version of rpm.  This is no different
> > from specifying an interface which corresponds to a particular version
> > of a kernel or C library.
> 
> The LSB should specify the binary install format used by a specific version
> of RPM. The source format and spec format might make sense to specify in
> an advisory capacity ditto the RPM command.
> 
> It doesnt matter how the rpm was built or what installed it. It matters that
> it installs everywhere
> 

The "rpm" command -- *or some subset thereof* might be specifiable if is
considered that this capability is important enough -- personally, I
would say it is, since otherwise there would be no way for a script to
install a package, and no way for a file browser to know how to install
a package.  However, there should be no requirement that the command
"rpm" invokes the program from Red Hat software.  If it is a script that
wraps "dpkg", then fine :)

This is very similar to what the POSIX committees often do: they specify
a set of features that are the minimum.  For example, I think "rpm -i",
"rpm -U", "rpm -F" and "rpm -e" (install and uninstall) are important
enough that they should be standardized, but "rpm -b" (build, i.e.
package creation) most likely is *not*.

Yes, this is work :)

	-hpa

-- 
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!