Updates in svn

[Kel Modderman]

Loïc Minier wrote:
>
>
>  I take the opportunity to request more information on the state of the
>  -ng branch: is it stable enough for day to day use?  known regressions?
>  it has been a while that I wonder whether it's in some useful state for
>  experimental, eg. I have no idea of the level of support for WPA/WPA2
>  of this branch, nor on the dependencies on external stuff (kernel level
>  and wpasupplicant for example).
>    Could you sched some light on this?  Is it now time for an upload?
>
>     Bye

The state of affairs with madwifi-ng as I see it is:

    * its stable on uniprocessor kernels for normal usage in
      station/managed mode or ap/master mode
    * there are no known arch specific problems
    * however, there are some locking issues on smp systems
    * most people report problems when really pushing the limits of the
      virtual device stuff (VAP's), especially a combination of
      different type virtual devices, for example an ap mode and sta
      mode combo
    * the frequency of reported problems have dramatically decreased in
      the last couple of months
    * wpasupplicant would require an update to include support for
      madwifi-ng, and that would not *not* be compatible with a madwifi
      interface using older drivers, however is am co-maintainer of
      wpasupplicant, and have been modifying it locally to support
      madwifi-ng for a long time
    * same situation for hostapd, although that maintainer has already
      included madwifi-ng headers in spite of the madwifi version
      actually in debian. To make things worse, the headers he includes
      have changed in upstream madwifi . . . It would also require an
      update to work 100% with current madwifi-ng
    * kismet would also require a svn patch/update to work with
      madwifi-ng, although, i believe kismet has planned to better
      support madwifi-ng in the near future (maybe a new release too)


It'd be absolutely fantastic to upload it to experimental (right now 
even!), however, it is *not* a suitable candidate for unstable at this 
point in time. Not until Michael Renzmann declares it suitable for mass 
consumption and makes a 1.0 release, or at least until after etch. And I 
am sure you already think the same.

I am using it daily (in ap and sta modes, with hostpad and 
wpa_supplicant), Matt Brown recently used madwifi-ng at linuxconf I 
believe, as the backbone of the wifi access points at the conference, 
and he continues to experiment with it in various exotic modes of operation.

The packaging of it is already tried and tested.

Matt Brown, Alex, would you also like to comment on this?

Thanks, Kel.