[conspire] "madwifi" is proprietary sludge (was: driver)

[Daniel Gimpelevich]

On Wed, 28 Jun 2006 02:03:29 -0700, Rick Moen wrote:

> Quoting Daniel Gimpelevich (daniel at gimpelevich.san-francisco.ca.us):
> 
>> There is that, and there is also the sad state of affairs of the
>> open-source drivers' support of the currently available hardware in that
>> category, as touched upon in Tim's post.
> 
> I don't know if it's quite _that_ dire -- though, as a habitual coward

No, it's really more dire than that.

> about hardware-related problems, I'm carefully sitting out the
> situation, and relying on my good ol' Lucent 802.11b chipsets, instead.
> 
> (Let us not forget that None of the Above is another time-honoured
> option.  See also "overdefining", below.)

Meh, depends on to whose time you're referring.

>> You also do not appear to make a distinction between a software limitation
>> and a hardware limitation. It is reasonable to expect software to overcome
>> a software limitation. It is not reasonable to expect software to overcome
>> a hardware limitation. 
> 
> The relevance of this would be clearer if you'd stated where I indicated
> any such expectation.

You did not, which was part of my point. It _is_ reasonable to expect
software to overcome a software limitation.

>> A software licensing limitation is a software limitation. If the
>> hardware may be used without the software, e.g.  using a NIC without
>> the CD it came with, the licensing limitation of said software is
>> thereby overcome. That is often not possible in the short term (stage
>> 1), but possible in the long term (stage 2).
> 
> It may be simple fatigue speaking but I'm not sure (1) exactly where I
> failed to make a distinction, or (2) what the relevance is to preceding
> discussion.

1) I was saying that ipw2200-type hardware has not-always-tolerable
hardware limitations, and you responded that Atheros hardware has a much
bigger limitation, which happens to be a software one.

2) We _were_ discussing the current and future status of Atheros hardware,
if I'm not mistaken.

>> The longer the term, the greater the percentage of hardware that has
>> reached stage 2, at which point, the comparison becomes moot.
> 
> If you're just saying "When nearly all choices become open source, then 
> open source ceases to be a comparative advantage", then it's obviously
> true but doesn't seem exactly profound.

Although that is also true, what I was trying to say was that given more
time, any specific stage 1 hardware gets more opportunity to reach stage
2, at which point it ceases to be a comparative advantage over that same
specific hardware.

>> > Speaking just for myself (and not, to acknowlege your intent, speaking
>> > for the broader Linux market), I never "overlook" proprietary software's
>> > short-term advantages.  I just note its long-term unacceptability (in my
>> > view), especially where tolerable alternatives exist already.
>> 
>> You define it as good in the short term and bad in the long term.
> 
> "Define" doesn't seem quite applicable in this context, but let's move
> on.  Also, I'm not sure "good" is quite right, there, either, but perhaps 
> "difficult to avoid" is closer.

Yes, I was trying to be succinct and sacrificed some accuracy there.

>> When the short term is stage 1 and the long term is stage 2, that
>> comes across as rather sdrawkcab. It would be great if there were
>> "tolerable alternatives" among 802.11g chipsets, but there are none at
>> this time.
> 
> 1.  My Lucent 802.11b cards do everything *I* need in a wireless chip.
> Methinks you're overdefining "tolerable alternative" for polemical
> reasons.

I specifically said "among 802.11g chipsets" here, and you pretty much
confirm that by having to go outside that category to find a tolerable
alternative. Your choice, the Orinoco a.k.a. Hermes I a.k.a. Prism I
chipset, has been supported by several drivers in the past, both
open-source and proprietary, which have all largely become abandonware,
with the sole exception of the "orinoco" driver. The development of that
driver has been very slow going, and much of it remains very kludgey. It
is still missing basic connectivity wherever authentication is required.
The most tolerable alternative among 802.11b chipsets are the Prism
2/2.5/3 chipsets, which are supported by several different open-source
drivers, including orinoco. Nobody uses the orinoco driver for them
because that would be like swapping the engine in a Lamborghini with a
spare one for a Yugo.

> 2.  I haven't seen you explain why a Prism 802.11g based card shouldn't
> be a reasonable choice.  (I don't know if they're made any more, but 
> that really wouldn't matter to me, either.  My Lucents were out of
> production when I bought them, too.  Es macht nicht.

I didn't see the need to explain that, since Tim illustrated such a
perfect example of why. The open-source driver that made the Prism
chipsets such an early favorite proved not to "be adaptable to meet
changing requirements and changing hardware" to the extent that it had to
be scrapped in favor of a from-scratch effort to re-support the hardware.
Hardware supported by the old, bit-rotting driver is increasingly hard to
find and not yet supported by the new driver, which is not yet mature nor
stable.

>> The proprietary HAL has a high degree of long-term unacceptability,
>> which ironically the Atheros chipset consequently lacks.
> 
> Huh?  I'm afraid I don't follow.  The only thing that makes the
> proprietary Atheros HAL not a problem in the long term is the recent
> emergence of the reverse-engineered OpenBSD alternative.  I hope you're
> not suggesting that Atheros Technology, with its deceptive press
> releases and all, can claim any credit for that.  Even crediting the
> almost equally misleading Madwifi Project is a bit of a stretch.

I'm having a little trouble determining whether or not I'm saying
something you hope I'm not suggesting, so I'll elaborate: At this time,
the most tolerable 802.11g chipset is Atheros, which leads to more
widespread proliferation of the hardware among people for whom the
long-term unacceptability of the proprietary HAL is an itch the OpenBSD
solution is intended to scratch, thereby feeding that development and
dispelling the long-term unacceptability of the hardware.

>> Exactly. Advances tomorrow will change what is advantageous or acceptable
>> tomorrow. However, what is advantageous or acceptable today depends both
>> on what is available today and on what will be available tomorrow.
> 
> I hope you don't think this insight is new to me.  ;->  I've kind of
> been at this hardware and Linux thing for a rather long time.  Please
> don't mistake my sometimes not knowing much about the latest toys with 
> ignorance of the industry:  It's just me being a tightwad.  

I thought I was merely restating what you said with the "tomorrow" bit,
and using it to further my point with the "today" bit.

> [Flash:]
> 
>> Not quite sure I follow. Coders had an itch prior to six years ago.
>> Suddenly, it didn't itch as much.
> 
> To explain:  The open source coders _never_ had much of an itch:  The
> historical lack of an open-source Flash interpreter not only wasn't a
> problem in the same league as lack of support for key hardware; in the
> opinion of many including yr. humble correspondent, it was practically a
> positive benefit!  Flash is mostly a blight on the Web, after all.

It may not have been much of an itch, but the itch did exist, and largely
disappeared with Macromedia's initial release. The forces that facilitated
that disappearance make the development effort of the open-source HAL
slower than it otherwise would be.

>> My point here was that Atheros support isn't itching as much as it
>> otherwise should.
> 
> Gods, I hope you're not _merely_ saying "Many people are OK with a
> short-term-functional proprietary, binary-only hardware driver, the
> existence of which will tend to reduce the urgency that will be
> perceived in many quarters for coding of an open-source one."  If you
> don't mind my saying so, that's bloody flippin' obvious -- and entirely
> irrelevant to my point.  

I was pointing out that it compounds the situation that the hardware is
already the most tolerable in spite of its stage 1 status, thus making
this bicycle-law anomaly slightly even more anomalous.

> To review:  You said slowness to develop good open-source Atheros
> drivers is comparable to slowness to develop good open-source Flash
> players after Macromedia's proprietary release for Linux.  In reply, I
> said, no, not really -- because of the rather vital difference that
> hardly anybody in open source gives a rat's ass about Flash regardless
> of licensing, except to loathe it, whereas hardware support is quite
> important to those same people.

Oh sure, ignore the similarities and concentrate on the differences...

> Some effect from Atheros/Madwifi dropping their half-a-loaf on the Linux
> open source coder community may or not have occurred (and I honestly
> don't think they're quite that gullible), but you've completely missed
> my point that you were drawing a parallel between something that
> matters and something that doesn't.

Not missed, merely dismissed as irrelevant...

>> Note that FlashBlock doesn't get triggered without your say-so, either
>> (i.e. your installation of it). 
> 
> Sorry, but your point completely eludes me.  Probably my fault,
> seriously.

This point was very tangential, and I'm not sure how much more clear I can
make it.

>> ATI has a history of contributing code to open-source drivers for hardware
>> that their crappy proprietary drivers no longer support. The history of
>> nVidia is one of zero open-source cooperation, coupled with relatively
>> meticulous maintenance of their proprietary drivers. I see this as ATI
>> doing more for the community than nVidia, but the prevailing
>> interpretation is that this means nVidia does more for the Linux
>> community. Clearly, ATI is the tolerable alternative here.
> 
> You are certainly correct that ATI has not been a total zero as a
> company.  However, the tolerable alternatives are, in fact, Matrox and
> Intel 945G (the latter if you give a rat's ass about 3D, which I do not).

ATI, as supported by the open-source drivers, is more tolerable than
either Matrox or Intel. The Intel driver isn't quite open-source, and
neither are the drivers for anything comparable from Matrox.

>> In the case of Celine Dion, preferred to what?
> 
> Blessed silence?  ;->

I gather that would sound somewhat like one hand clapping...