Re: [Madwifi-devel] Usespace API/ABI

[Pavel R. <pr...@gn...>]

Quoting Derek Smithies <de...@in...>:

> But we do not distribute the iwTools. Thus, changes to the output of
> commands like iwpriv etc are not good - they make life a bit harder for
> end user.
>  In particular, any scripts written which process the output of iwpriv
> require fixing.

I agree.  I was talking about the interfaces available only to other utilities,
not to the end user.

As you can see, there is no exact boundary between a program interface and a
user interface.  And then we need to further distinguish interfaces to programs
distributed with MadWifi and to programs distributed separately.

One thing I'd like to do is to make wpa_supplicant and hostapd compile for
MadWifi without MadWifi headers.  For this to be reliable, the ioctl numbers
could be requested by name using wireless extensions.  For hostapd interface,
MadWifi could emulate the Devicescape interface, known as cfg80211.  Once this
is implemented, the old intefaces could be phased out.

The ability to support MadWifi without the MadWifi source would make it easy to
enable MadWifi support by default in some distributions, which is a good thing
and could be a good reason to change some interfaces if required.

Arguing about a general policy is probably not the best approach.  We need to
know what every particular change would achieve and what it would break.

--
Regards,
Pavel Roskin