Ksenija wrote:
> I'm helping Greg do a bit of cleanup in the staging tree, I noticed that
> wlan-ng driver is maybe ready to be moved out of staging. Are there
> any TODO tasks left to do beside checkpatch.pl clean-up?
Its FAQ was very clear: ftp://ftp.linux-wlan.org/pub/linux-wlan-ng/FAQ
[archived at http://puppylinux.dreamhosters.com/wireless/linux-wlan-org_FAQ.txt ]
--cut--
Q: When will linux-wlan-ng be merged into the mainline kernel?
Short answer: Never.
First, the linux-netdev people will soundly reject this driver.
I don't begrudge them for this; indeed in their position I'd do
exactly the same. It's a sound engineering decision.
linux-wlan-ng is obselete, and effort spent fixing it is better
spent elsewhere. You can't even buy the hardware any longer.
The original design for linux-wlan-ng was to separate the 802.11
stack from the actual hardware driver. This added a lot of
complexity, but would greatly ease the pain of supporitng
multiple hardware types. Unfortunately, the implementation was
turned out to be somewhat flawed, and hardware manufaturers went
away from the thick-mac model, leaving linux-wlan-ng overly
complex for what it did.
(Ironically, the linux kernel is adopting a similar
separation model, but it is a long way off from being ready)
So why not rewrite linux-wlan-ng to be more suitable, the
enterprising reader may ask?
The kernel already has two drivers for prism2 (cs/pci/plx)
hardware -- hostap and orinoco. linux-wlan-ng basically
has three features not present in kernel drivers:
1) USB support
2) nearly complete implemettion of the 802.11 MIB/MLME
3) Firmware-based AP support
(3) requires an expensive license that isn't even available any
longer, as the hostap mode works far better -- and is already
supported by in-kernel drivers.
(2) would need to be removed or completely rewritten in order to
be merged, as it does not fit within existing kernel APIs, and
it would be effectively merging new kernel APIs.
(1) Is the only truly unique thing that linux-wlan-ng does that
is generally needed any more.
To merge it into the kernel, we'd need to strip out (2), which
would necessitate a complete rewrite -- to the point where
writing a new driver from scratch is easier.
Basically, it would take far less effort to add USB
support to the in-kernel drivers than it would to make
linux-wlan-ng acceptable to be merged.
In other words, the short answer is: Never.
--end--
On Fri, Jan 29, 2016 at 05:00:16PM +0100, Xose Vazquez Perez wrote:
> Ksenija wrote:
>
> > I'm helping Greg do a bit of cleanup in the staging tree, I noticed that
> > wlan-ng driver is maybe ready to be moved out of staging. Are there
> > any TODO tasks left to do beside checkpatch.pl clean-up?
>
> Its FAQ was very clear: ftp://ftp.linux-wlan.org/pub/linux-wlan-ng/FAQ
> [archived at http://puppylinux.dreamhosters.com/wireless/linux-wlan-org_FAQ.txt ]
>
> --cut--
> Q: When will linux-wlan-ng be merged into the mainline kernel?
>
> Short answer: Never.
The ultimate aim of a driver in staging is to graduate out of staging
area into the main part of the kernel. If it is already known that this
driver can never move out of staging then IMHO its better to remove it.
regards
sudip
On Fri, Jan 29, 2016 at 05:11:04PM +0100, Xose Vazquez Perez wrote:
> >> I'm helping Greg do a bit of cleanup in the staging tree, I noticed that
> >> wlan-ng driver is maybe ready to be moved out of staging. Are there
> >> any TODO tasks left to do beside checkpatch.pl clean-up?
> >
> > Its FAQ was very clear: ftp://ftp.linux-wlan.org/pub/linux-wlan-ng/FAQ
> > [archived at http://puppylinux.dreamhosters.com/wireless/linux-wlan-org_FAQ.txt ]
As author of that particular FAQ entry (written nearly ten years ago)
let me say that nothing in there has any bearing on the situation that
exists today.
At the time, I didn't think it was worth the significant effort to get
prism2_usb into the mainline, but obviously some enterprising folks
disagreed with me, and eventually put forth the elbow grease necessary
to get it into -staging and all of the subsequent cleanups.
From where I sit, the basic question is this: Is the prism2_usb driver
now considered good enough to promote out of -staging? If not, what
needs doing and who is willing to see it through? If it is, then just
do it already. :)
Oh, I'm still hosting the firmware files needed for this thing, and I
see a consistent 4K or so downloads a month. So there are obviously
users out there still.
- Solomon
--
Solomon Peachy pizza at shaftnet dot org
Delray Beach, FL ^^ (email/xmpp) ^^
Quidquid latine dictum sit, altum viditur.
On 01/29/2016 05:00 PM, Xose Vazquez Perez wrote:
> Ksenija wrote:
>
>> I'm helping Greg do a bit of cleanup in the staging tree, I noticed that
>> wlan-ng driver is maybe ready to be moved out of staging. Are there
>> any TODO tasks left to do beside checkpatch.pl clean-up?
>
> Its FAQ was very clear: ftp://ftp.linux-wlan.org/pub/linux-wlan-ng/FAQ
> [archived at http://puppylinux.dreamhosters.com/wireless/linux-wlan-org_FAQ.txt ]
>
> --cut--
> Q: When will linux-wlan-ng be merged into the mainline kernel?
>
> Short answer: Never.
>
> First, the linux-netdev people will soundly reject this driver.
> I don't begrudge them for this; indeed in their position I'd do
> exactly the same. It's a sound engineering decision.
>
> linux-wlan-ng is obselete, and effort spent fixing it is better
> spent elsewhere. You can't even buy the hardware any longer.
>
> The original design for linux-wlan-ng was to separate the 802.11
> stack from the actual hardware driver. This added a lot of
> complexity, but would greatly ease the pain of supporitng
> multiple hardware types. Unfortunately, the implementation was
> turned out to be somewhat flawed, and hardware manufaturers went
> away from the thick-mac model, leaving linux-wlan-ng overly
> complex for what it did.
>
> (Ironically, the linux kernel is adopting a similar
> separation model, but it is a long way off from being ready)
>
> So why not rewrite linux-wlan-ng to be more suitable, the
> enterprising reader may ask?
>
> The kernel already has two drivers for prism2 (cs/pci/plx)
> hardware -- hostap and orinoco. linux-wlan-ng basically
> has three features not present in kernel drivers:
>
> 1) USB support
> 2) nearly complete implemettion of the 802.11 MIB/MLME
> 3) Firmware-based AP support
>
> (3) requires an expensive license that isn't even available any
> longer, as the hostap mode works far better -- and is already
> supported by in-kernel drivers.
>
> (2) would need to be removed or completely rewritten in order to
> be merged, as it does not fit within existing kernel APIs, and
> it would be effectively merging new kernel APIs.
>
> (1) Is the only truly unique thing that linux-wlan-ng does that
> is generally needed any more.
>
> To merge it into the kernel, we'd need to strip out (2), which
> would necessitate a complete rewrite -- to the point where
> writing a new driver from scratch is easier.
>
> Basically, it would take far less effort to add USB
> support to the in-kernel drivers than it would to make
> linux-wlan-ng acceptable to be merged.
>
> In other words, the short answer is: Never.
> --end--
>
Detailed info in these threads:
https://marc.info/?l=linux-wlan-user&m=125122113309475
https://marc.info/?l=linux-wireless&m=138384105903345