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

[Rick Moen]

Quoting Daniel Gimpelevich (daniel at gimpelevich.san-francisco.ca.us):

> It's not just that; there are many other factors contributing to this
> slowness. First, until ipw2200, there was no stage 2 802.11g chipset.
> Intersil came close, but did not reach the cigar.

I'm curious about why you say that about Prism/Intersil, because for
quite a long time I guessed that it (Prism whatever) is the one I'd buy,
if/when I got around to replacing my Lucent 802.11b stuff.  It certainly
looks to me like a perfect example of stage 2, with only the relatively
trivial objection that Intersil / Globespanvirata / Conexant never
_technically_ issued permission for the public to redistribute its BLOB
files.  (So, the various sites redistributing them are committing
technical copyright violation, but are comfortably in the "fortunately,
nobody cares" category.)

Those BLOB files, as noted, are firmware equivalents, only, and have no
OS-specific contents.  (I don't especially mind my Lucent's ROM code 
being proprietary because it's not general-purpose software.  I don't
especially mind that trait in the Intersil BLOB because it fills exactly
the same role.)

> So other than temporarily making Intersil hardware a prized commodity,
> prices remained what they were, where ipw2200-type hardware was the
> cheapest. Now that the ipw2200 driver is mature, the hardware it
> supports is known more for its limitations than for that support.

I notice we have a difference of perception on weighting:  I consider a key
piece of system driver software being proprietary (and especially of it 
being binary-only) as being the most serious kind of "limitation" of
all:  It guarantees that the software will not be adaptable to meet
changing requirements and changing hardware, and ensures that it will 
cease to be maintainable at all, the moment its owner loses interest or
ability in doing so.  It becomes abandonware.

This is one of the chief pragmatic reasons for open source being an
operational advantage over the longer term.  (There are of course other
sorts of reasons.)


> Add to that the existence of the madwifi driver with its HAL, which
> leads people to overlook the stage 1 status of Atheros hardware
> because it fully works.

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.

Over the longer term, those alternatives are likely to (themselves)
improve:  Any advances (in it or even _similar_ projects for different
but related hardware) become permanently available to everyone, not held
hostage to the fortunes and intentions of a copyright owner.  By the
same token, the proprietary software is, by contrast, doomed in the long
term by the abandonware problem.


> I would compare this to the evaporation of nearly all efforts to
> fabricate a fully functional open-source SWF player six years ago when
> Macromedia released their first Flash player for Linux.

Actually, I see a big difference:  Nobody's connectivity relies on a
Flash interpreter, and it's largely mandated by advertising and online
cartoons.  In other words, potentially motivatible coders, by and large,
don't actually have an itch they're impelled to scratch.

Personally, I go rather far out of my way to _not_ have a Flash
interpreter, and to make sure any one that does happen to be present
never gets triggered without my say-so.  (I.e., I install FlashBlock.)


> Tim has just posted a perfect example of why Atheros is still the
> preferred 802.11g chipset for Linux: because of the research
> beforehand, not because of the lack of it.

...by those who haven't yet learned the lesson about avoiding getting
trapped into proprietary software ghettos unnecessarily.  See also: 
NVidia and ATI proprietary X11 drivers.

(By "preferred", you probably mean nothing beyond "popular".  Lots of
appalling things are that, up to and including Celine Dion.  ;->  )


> So, here we have an instance where the value of stage 2 hardware
> continues to be at a low point, while the value of stage 1 hardware is
> improved.

Listen, I didn't say that the natural economics of demand for
open-source drivers cannot be overridden by (1) an overwhelming number
of ignorant, short-sighted customers and (2) pigheaded manufacturers,
some of whom brazenly lie about their drivers' nature and claim to 
be doing open source when they aren't.  (In fact, that's what my link to
Moen's Law of Bicycles addresssed, neh?)