Received-SPF: pass (google.com: domain of linux-wireless-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210820043538.12424-1-pkshih@realtek.com> <20210820043538.12424-3-pkshih@realtek.com>
 <87sfxkix5e.fsf@codeaurora.org>
In-Reply-To: <87sfxkix5e.fsf@codeaurora.org>
From:   Brian Norris <briannorris@chromium.org>
Date:   Fri, 1 Oct 2021 10:40:39 -0700
Message-ID: <CA+ASDXMX2Kw9JLBtzAvEJr=qFD9rq4L6+zFeHeYCPUeOc=u7pw@mail.gmail.com>
Subject: Small driver submissions and long feedback cycles
To:     Kalle Valo <kvalo@codeaurora.org>
Cc:     Ping-Ke Shih <pkshih@realtek.com>,
        linux-wireless <linux-wireless@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

[[ changing subject, as this is less about rtw89; it used to be:
  Re: [PATCH v6 02/24] rtw89: add BT coexistence files ]]

On Fri, Oct 1, 2021 at 8:26 AM Kalle Valo <kvalo@codeaurora.org> wrote:
> A small tip for future drivers, try to remove all the optional features
> from the driver as much possible and keep only the absolutely needed
> features to get ping working. For example this file was pain to review
> and I suspect coex support could have been submitted separately.

I find that a bit of a tall order. You haven't looked at this driver
for 9-10 months:

https://lore.kernel.org/linux-wireless/20201230044223.14085-1-pkshih@realtek.com/

Do you expect people to just live with a feature-less driver (read:
unusable) for all that time? As noted in other parts of this thread,
quite a few people are already using this driver; so your suggestion
implies people should submit a completely different driver to you
(i.e., more or less non-working) than the one everybody else wants to
use -- or else suffer for those 9-10 months.

For the record, Realtek _did_ try this for rtw88, where they
partitioned their driver work into stages, submitting you only the
first stuff over the period of 9 months before you merged it:

rtw88 RFC v2; I couldn't find v1:
https://lore.kernel.org/linux-wireless/1538553748-26364-1-git-send-email-yhchuang@realtek.com/
Committed about 9 months later:
https://git.kernel.org/linus/e3037485c68ec1a299ff41160d8fedbd4abc29b9

But it was several months after that before the rest of the
really-usable features were submitted and actually landed properly.

So it sounds like with your suggested approach (like rtw88), it can
take over a year to get a usable driver merged. For this approach, I
guess it's 9+ months (TBD; but you seemed more or less happy with this
version, minus some small comments). I can see why folks would choose
the latter.

Or maybe, those timelines sound altogether bad [1], and we should
consider some alternative. If the review and feedback (or merge) cycle
was quicker, I'm sure people would be happier to split work into
bite-sized chunks. But when it's slow and lacking in transparency,
people are instead incentivized to just publish everything at once. Do
you need additional help? A reviewer team? I do occasionally try to
help things a lot with review on-list, but it's not clear that it has
any bearing on time-to-acceptance, so I don't exactly stretch myself.

Brian

[1] They do to me.