HEADS-UP: removed COMPAT_43 from GENERIC (and other configs)

[Divacky Roman]

On Fri, Jun 16, 2006 at 01:29:14PM -0600, Scott Long wrote:
> Poul-Henning Kamp wrote:
> 
> >In message <4491C2F0.6000007 at rogers.com>, Mike Jakubik writes:
> >
> >
> >>What about COMPAT_43TTY? Is this still needed, how exactly does it 
> >>affect the system?
> >
> >
> >It adds a bunch of ancient-compatible ioctls to the kernel.
> >
> >It is, as a principle, not needed, but thanks to the many variants
> >of "sh configure" employed in usr/ports, a quite large number of
> >ports go "Ohh, this is BSD, I'd better use the old ioctls" and
> >break if you don't offer them.
> >
> 
> One thing to keep in mind is that upgrade compatibility is very
> important.  Not everyone lives at the tip of the tree, and not
> everyone wants to, or even can, recompile all of their apps for
> an upgrade.  Making COMPAT_43 and COMPAT_43TTY be optional is fine,
> and fixing as many ports as possible not to rely on it is fine too,
> but removing the options from the kernel will be a mistake right now.
> People were running 2.2.x apps well into the 4.x lifecycle, and people
> are running 4.x apps now well into the 6.x lifecycle.  If you make
> their lives harder, you'll make it a lot easier to justify switching
> to something else.  If you want to deprecate and ultimately removethese
> options, set a 2-3 year timeline for it, and heavily advertise it.
> Anything shorter will do more harm than good.

while I tend to agree you have to see that COMPAT_43 ensures in-kernel
compatibility layer so there should be no (user-space) app breakage. and noone
is removing COMPAT_43TTY now

roman