Re: WPA status 2006-01-08

[Dan Williams <dcbw redhat com>]

On Mon, 2006-01-09 at 12:27 -0500, Robert Love wrote:
> On Mon, 2006-01-09 at 12:18 -0500, Dan Williams wrote:
> 
> > Note that while wpa_supplicant supports using driver-specific methods
> > for WPA and other settings, we want to push all drivers towards
> > conforming to the WEXT spec on this one.  That means support for
> > SIOCSIWAUTH and SIOCSIWENCODEEXT.  We _may_ have to allow
> > driver-specific support in the mean time, but I'd rather not do that if
> > at all possible.  (for instance, atmel doesn't seem to work right now
> > for normal WEP)
> 
> We are going to need to go through the various drivers and see how they
> fair.  We are probably going to neeed driver-specific support.

Even if that's the case, we're going to need to push those drivers
towards WEXT compliance, such that they do what they need to do with the
wpa_supplicant "wext" driver.  I'm much more amenable to making sure
they all work with WEP & wpa_supplicant first, and taking more time with
WPA.

For example, the "atmel" driver for wpa_supplicant doesn't work on the
in-kernel atmel driver _at__all_, probably because it expects
atmelwlandriver.sf.net rather than the in-kernel one.  I'm looking at
fixing that up for WEP-only at the moment.

But unfortunately we do have some regressions right now, and we've got
to look at how to fix those.  If we do go driver-specific in
NetworkManager, then there really will be a Flag Day where we turn off
that support and force drivers to be WEXT compliant.  If distros don't
like that, they can either fix the drivers or patch NM (Fedora
included).  I'd like that day to be as soon as realistically possible.

Dan