Received-SPF: pass (google.com: best guess record for domain of linux-wireless-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <CAAooHFeLWrY_wmCp-HWqygh8gnKsfpoPCky7SykOBKZgXkb8OQ@mail.gmail.com>
 <20190403210200.GA93453@google.com> <211816ff03cf188d834a21b1fbc98b4f8c5b81f4.camel@sipsolutions.net>
 <CA+ASDXOyXb0dPGOrjQR7C-b6dyftiZhkta3cwG28B9sC5wxHxQ@mail.gmail.com> <7687225C-D965-479E-BAE8-769B0AEADD76@holtmann.org>
In-Reply-To: <7687225C-D965-479E-BAE8-769B0AEADD76@holtmann.org>
From:   Brian Norris <briannorris@chromium.org>
Date:   Fri, 16 Aug 2019 13:29:43 -0700
Message-ID: <CA+ASDXNC0hwFzSTvZmUq-B7r_H+pZ3N=p_kjfMqKb1gftsmDKw@mail.gmail.com>
Subject: Re: Flag for detecting 802.11r Fast BSS Transition support
To:     Marcel Holtmann <marcel@holtmann.org>
Cc:     Johannes Berg <johannes@sipsolutions.net>,
        Matthew Wang <matthewmwang@google.com>,
        linux-wireless <linux-wireless@vger.kernel.org>,
        Kirtika Ruchandani <kirtika@google.com>,
        Jouni Malinen <j@w1.fi>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-wireless-owner@vger.kernel.org
Precedence: bulk

Hi Marcel,

On Fri, Aug 16, 2019 at 11:54 AM Marcel Holtmann <marcel@holtmann.org> wrote:
> > Well, I guess we could just run the command and look for EOPNOTSUPP...
>
> this kind of API design and usage is bad. Try-and-error approach is just not sustainable.

Sure. That "suggestion" was quite literally an afterthought. Not
really a proper suggestion.

> Even while it is late to add a proper flag that indicates support, we need to do this to make nl80211 better for the future.

I suppose. I'm not quite sure how I would make use of that properly
though, given the corpus of kernels out there where the flag doesn't
exist (but the feature does). Some other heurestic for determining
kernel recency? Compile-time flags for the relevant user space, such
that one builds it for "new kernel API (w/ flag)" vs. "old kernel API"
(with the latter not even trying to look for the flag)?

Or I guess a more proactive approach: implement both a "supported" and
an "unsupported" flag, so user space can figure out a tristate: flag
not available (old kernel -- user space is left to guess) vs. command
supported flag vs. command not supported flag.

That seems a bit awkward though.

Brian