Re: PPP problem w/ 1.1 install
On Fri, 31 May 1996, Ian Jackson wrote:
> Manoj Srivastava writes [ SuperCite undone - iwj ]:
> > Brian C White <email@example.com> writes:
> > > So... Should there be a restriction against listing the
> > > kernel-image as a dependancy in another package?
> > No, since if you follow the recommended method of generating
> > kernel images, this will work.
> On the contrary, we should not require people to follow this method,
> especially when it's easy not to make this requirement.
> Packages which need a particular kernel or kernel feature to run
> correctly should test for the kernel version of feature in the
> postinst (or in the preinst, if the package being broken is a serious
> problem for the whole syste, for example for a base package).
This fails to provide protection if the kernel is downgraded to a kernel
that never provided the desired feature, or upgrading to a kernel that
provides this feature in a different manner or with a different interface.
I would suggest that it is always inappropriate for a package to depend on
a particular kernel version. It is much better for the package in question
to depend on a version of libc that provides the desired interface to the
kernel. Then the kernel version can change without breaking the
application program. It was my understanding that this is exactly why
David E. began providing kernel headers with libc5.
I say: Use depends, but make it depend on the appropriate version of libc
and not on a particular kernel package.
aka Dale Scheetz Phone: 1 (904) 877-0257
Flexible Software Fax: NONE
Black Creek Critters e-mail: firstname.lastname@example.org
------------ If you don't see what you want, just ask --------------