From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <5b596a6460a357d420e1d176d573ecd4@quanstro.net>
From: erik quanstrom <quanstro@quanstro.net>
Date: Mon, 19 Jan 2009 11:43:07 -0500
To: 9fans@9fans.net
In-Reply-To: <13426df10901190825x2a6d79c5g2b787de2617f9fc1@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: Re: [9fans] Les Mis?rables
Topicbox-Message-UUID: 82c23c56-ead4-11e9-9d60-3106f5b1d025

> drivers should be easy. What turns them hard is dealing with all the
> bugs in the hardware.

vendors vary widly in the amount of bugs there are to
deal with.  errata can be your friends.  i can't speak
to wireless parts, but my experience is that intel parts
generally work quite well with few errata.

however, the errata are not necessarly as bad as the
odd features!

i have a recent experience.  a certain $vendor has a
$part.  it has some bits in a register that specify the
"endianness" of the data.  unfortunately, the $spec
specifies the bytes in order.  it took a good long while
to figure out that the $part's docs were specifying
bytes with respect to the native word byte order.
thus the mapping of the bits was the reverse of the
expected.

why such a feature would be included is beyond me.
the special bits need to be fiddled together regardless
and the data have no ordering.

- erik