[lsb-discuss] Question from an AppChecker user...

[Theodore Tso]

On Fri, Dec 12, 2008 at 11:52:35AM -0700, Wichmann, Mats D wrote:
> As far as I know, AppChecker has boxes to submit "errors" as
> "request for future LSB feature" - I haven't looked at this in
> detail but predecessor tools (atk-manager) had this.
> 
> I'm assuming from the rest of the context "kernel API" means an
> interface for drivers to use.  If such a thing truly stabilizes,
> then someone could propose to include it; while LSB has been asked
> for it many many times, I can say as an internal statement - maybe
> this is not what you pass on - that LSB is not working on that with
> anyone right now.

Upstream has already made its statement on this matter.  See the
stable_api_nonsense.txt in the Linux kernel sources:

http://www.kernel.org/pub/linux/kernel/people/gregkh/misc/2.6/stable-api-nonsense-2.6.10-rc2.patch

> I think dkms is considered an interim solution until other things
> evolve (as discussed above).  I'd say if it's really considered
> interim it would be unlikely for LSB.

I'd have to agree; *highly* unlikely.

> This sounds negative; the essence is that LSB has been focused on a
> different layer of application portability than is implied by
> talking directly to the kernel.  I *know* there are applications
> that have a driver component, and those have always seemed
> problematic for the apps which have a driver piece.  It would take
> some sort of conscious decision for LSB to start focusing on that
> area, which doesn't mean it couldn't happen someday.

We've always said that the LSB won't standardizing anything without
the support of the upstream authors/maintainers; and while anything is
possible, including Greg K-H stating (and not as a joke) in a keynote,
"Bill Gates is my God and Ballmer is his prophet" ---- well, it just
doesn't seem highly likely.  :-)

							- Ted