[PATCH] mac80211: mark as experimental

[PATCH] mac80211: mark as experimental

[Jiri Benc]

Make the stack depend on EXPERIMENTAL. Also, change the name from "dscape"
to "mac80211".

Signed-off-by: Jiri Benc <jbenc@suse.cz>

---
 net/mac80211/Kconfig |    3 ++-
 1 files changed, 2 insertions(+), 1 deletion(-)

--- mac80211.orig/net/mac80211/Kconfig
+++ mac80211/net/mac80211/Kconfig
@@ -1,5 +1,6 @@
 config MAC80211
-	tristate "Generic IEEE 802.11 Networking Stack (dscape)"
+	tristate "Generic IEEE 802.11 Networking Stack (mac80211)"
+	depends on EXPERIMENTAL
 	select CRYPTO
 	select CRYPTO_ECB
 	select CRYPTO_ARC4



-- 
Jiri Benc
SUSE Labs

Re: [PATCH] mac80211: mark as experimental

[Pavel Roskin]

On Mon, 2007-04-30 at 20:17 +0200, Jiri Benc wrote:
> Make the stack depend on EXPERIMENTAL. Also, change the name from "dscape"
> to "mac80211".

Why is it needed to make mac80211 EXPERIMENTAL?  With mac80211 already
in Ubuntu 7.04 and in the forthcoming Fedora 7, it must have received
more testing than some drivers not marked as EXPERIMENTAL.

I don't think enabling mac80211 per se would destabilize the kernel.
Individual drivers could be marked experimental.  In particular,
bcm43xx_mac80211 needs such designation, since the softmac version is
much better currently.

If we encourage enabling mac80211, it would make it easier for the users
to compile and test mac80211 based drivers, such as iwlwifi.  Even
bcm43xx_mac80211 could be packaged separately to allow compilation
against an existing kernel for the users who want to test it without
upgrading the kernel.

In my opinion, marking mac80211 as EXPERIMENTAL would hinder testing of
mac80211 drivers without having any positive effect on the users who
want stability.

-- 
Regards,
Pavel Roskin

Re: [PATCH] mac80211: mark as experimental

[Dan Williams]

On Mon, 2007-04-30 at 17:48 -0400, Pavel Roskin wrote:
> On Mon, 2007-04-30 at 20:17 +0200, Jiri Benc wrote:
> > Make the stack depend on EXPERIMENTAL. Also, change the name from "dscape"
> > to "mac80211".
> 
> Why is it needed to make mac80211 EXPERIMENTAL?  With mac80211 already
> in Ubuntu 7.04 and in the forthcoming Fedora 7, it must have received
> more testing than some drivers not marked as EXPERIMENTAL.

It's still just not there yet, especially WRT compatibility behavior for
WEXT.  I don't know if that justifies EXPERIMENTAL status, but most
drivers don't appear to be production ready, from user reports.

Dan

> I don't think enabling mac80211 per se would destabilize the kernel.
> Individual drivers could be marked experimental.  In particular,
> bcm43xx_mac80211 needs such designation, since the softmac version is
> much better currently.
> 
> If we encourage enabling mac80211, it would make it easier for the users
> to compile and test mac80211 based drivers, such as iwlwifi.  Even
> bcm43xx_mac80211 could be packaged separately to allow compilation
> against an existing kernel for the users who want to test it without
> upgrading the kernel.
> 
> In my opinion, marking mac80211 as EXPERIMENTAL would hinder testing of
> mac80211 drivers without having any positive effect on the users who
> want stability.
> 

Re: [PATCH] mac80211: mark as experimental

[Pavel Roskin]

On Tue, 2007-05-01 at 07:51 -0400, Dan Williams wrote:
> On Mon, 2007-04-30 at 17:48 -0400, Pavel Roskin wrote:
> > On Mon, 2007-04-30 at 20:17 +0200, Jiri Benc wrote:
> > > Make the stack depend on EXPERIMENTAL. Also, change the name from "dscape"
> > > to "mac80211".
> > 
> > Why is it needed to make mac80211 EXPERIMENTAL?  With mac80211 already
> > in Ubuntu 7.04 and in the forthcoming Fedora 7, it must have received
> > more testing than some drivers not marked as EXPERIMENTAL.
> 
> It's still just not there yet, especially WRT compatibility behavior for
> WEXT.  I don't know if that justifies EXPERIMENTAL status, but most
> drivers don't appear to be production ready, from user reports.

That's my whole point.  Mark the drivers experimental, not the common
code.

I think the issues affecting WEXT in non-mac80211 drivers by cfg80211
were fixed in December or so.

But anyway, it's better to have "experimental" mac80211 in the kernel
than no mac80211 at all.

-- 
Regards,
Pavel Roskin