[pkg-wpa-devel] Removal of madwifi driver backend support

[Kel Modderman]

On Sunday 25 May 2008 22:03:41 Kel Modderman wrote:
> Hi,
> 
> I would like to propose the removal of madwifi specific backend from
> wpasupplicant package for the following reasons:
> 
> 1) The use of -Dmadwifi is actively warned against by upstream Madwifi
>    developers. wext should be used instead.
> 
> 2) If there are bugs in wext driver backend in conjunction with madwifi, they
>    should be reported, and they will be more likely to be fixed since wext is
>    the preferred standard of operation for both Madwifi and
>    hostap/wpa_supplicant upstream support and development personell.
> 
> 3) The use of a snapshot of madwifi include files to enable private interface
>    introduces ABI incompatibility when upstream Madwifi changes. There is no
>    facility to ensure a stable wpa_supplicant/madwifi backend interface with
>    various different versions of madwifi driver. The fact that this has not
>    changed in a while is pure luck; the Madwfi project's delivery of new binary
>    HAL + glue has stagnated for a long time now.
> 
> 4) The madwifi include files constitute a huge part of the debian difference to
>    upstream, making the package slightly more difficult to maintain.
> 
> 5) NetworkManager uses "wext" as backend for, as far as I can see after
>     inspecting source code, all wireless devices.
> 
> 6) Compatibility code could be put into ifupdown integration to display a
>    message when "wpa-driver madwifi" line was used in a /e/n/i stanza, and
>    the operation could continue with wext backend.

Please note that I managed to do this in r1170, but did not mean to. I repeat
that r1170 was a mistake.

Meanwhile, I spotted that driver_hostap is not needed by any hostap-driver
driven device using Linux 2.6.14 or above and can be disabled in favour of
wext.

Also, driver_test doesn't really make sense in a binary package, especially
without eap test code, therefore it was disabled.

Thanks, Kel.