On Thu, May 22, 2008 at 02:29:45PM -0700, Jordan Brown wrote:
> Stephen Hahn wrote:
> >   If that file goes outside the image, it's not really a packaging
> >   action.  So far, I've only heard an attempt to shift such operations
> >   back into packaging--why aren't these installer layer
> >   responsibilities?
> 
> I think the real answer is that the systems (SVR4, RPM) that you're 
> replacing don't *require* an installer layer.  Many applications find 
> that they must add an installer layer, either from laziness or from gaps 
> in the "packaging" layer, but I personally consider that to be a failure 
> rather than a model to emulate.

SVR4 provides some functionality that I think the IPS folks consider
installer functionality, but we certainly do layer an installer (or N)
on top of SVR4...  :^/

I don't mind such installer functionality being moved out of the
packaging system, as long as there's a generic component that provides
some such functionality.

> If you're going to require an installer layer, then it is the installer 
> layer that users should interact with, with the "packaging" layer being 
> an implementation detail hidden underneath it.

I'm not sure I agree.  I agree with Bart's note about not running
packaged code on a system that's installing into a non-live image.  And
that must be supported, so it seems like the installer functionality in
question must be a post-pkg install step.  Again, I don't mind that.

We'll see what happens.  As long as we get the ping fmri mechanism we
can build much of the rest as we need it; user images vis-a-vis
GNOME-like things may just have to deal.

Nico
-- 
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to