Open hardware wireless cards

On Wed, Jan 05, 2005 at 02:24:47PM -0500, Luis R. Rodriguez wrote:
<-- snip -->

As far as support for the new chipsets goes -- sorry -- we won't be able
to support it as I don't think even Conexant has a final well tested
linux source base ready for 2.6. And even if we are given a source base
there is nothing we can do to get around the need for the closed-source 
softmac libs that it relies on. As much as I'd like to support it, I
don't want to get a headache to support something I cannot modify so I
won't be willing to support a half-opened driver as the atheros driver.

I'd also like to add...

For those of you frustrated about our current wireless driver situation
in open platforms --

I think we probably will have this trouble with most modern hardware for a while
(graphics cards, wireless driver, etc). A lot of has to do with patent
infringement issues, "intellectual property" protection, and other
business-oriented excuses.

What I think we probably will have to do is just work torwards seeing if
we can come up with our own open wireless hardware. I know there was
a recent thread on lkml about an open video card -- anyone know where
that ended up?

If we can't come up with our own project to work on open hardware we can
also just see if its feasible to purchase hardware companies on the
verge of going backrupt and buy them out and release the specs/etc (a la
blender). Can someone do the math here? I'm lazy.

	Luis

-- 
GnuPG Key fingerprint = 113F B290 C6D2 0251 4D84  A34A 6ADD 4937 E20A 525E

On Wed, Jan 05, 2005 at 03:05:26PM -0500, Luis R. Rodriguez wrote:

I'd also like to add...

For those of you frustrated about our current wireless driver situation
in open platforms --

I think we probably will have this trouble with most modern hardware for a while
(graphics cards, wireless driver, etc). A lot of has to do with patent
infringement issues, "intellectual property" protection, and other
business-oriented excuses.

What I think we probably will have to do is just work torwards seeing if
we can come up with our own open wireless hardware. I know there was
a recent thread on lkml about an open video card -- anyone know where
that ended up?

If we can't come up with our own project to work on open hardware we can
also just see if its feasible to purchase hardware companies on the
verge of going backrupt and buy them out and release the specs/etc (a la
blender). Can someone do the math here? I'm lazy.

Being open doesn't mean you aren't violating some stupid patent.
Software patents really are an incredibly stupid idea.  Algorithms are
pushing it.  After all what does the algorithm do on it's own?  Can you
show it working and doing something?

Len Sorensen

On Wed, 5 Jan 2005 lsorense@csclub.uwaterloo.ca wrote:

On Wed, Jan 05, 2005 at 03:05:26PM -0500, Luis R. Rodriguez wrote:

quoted

I'd also like to add...

For those of you frustrated about our current wireless driver situation
in open platforms --

I think we probably will have this trouble with most modern hardware for a while
(graphics cards, wireless driver, etc). A lot of has to do with patent
infringement issues, "intellectual property" protection, and other
business-oriented excuses.

What I think we probably will have to do is just work torwards seeing if
we can come up with our own open wireless hardware. I know there was
a recent thread on lkml about an open video card -- anyone know where
that ended up?

If we can't come up with our own project to work on open hardware we can
also just see if its feasible to purchase hardware companies on the
verge of going backrupt and buy them out and release the specs/etc (a la
blender). Can someone do the math here? I'm lazy.

Being open doesn't mean you aren't violating some stupid patent.
Software patents really are an incredibly stupid idea.  Algorithms are
pushing it.  After all what does the algorithm do on it's own?  Can you
show it working and doing something?

Len Sorensen

Well its mostly about making boards just a bit too cheap.

If the vendors would just put in a serial EPROM that loads the
proprietary stuff to their PLD upon startup, then there is
no "proprietary" code to worry about. You have the generic
published interface like a PLX chip, plus the specific published
register functions of their PLD chip. How the device actually
makes smoke and mirrors is hidden even from the programmer.

But, by eliminating the US$0.50 cost of a EPROM, they want to supply
a sack of bits that needs to be uploaded to the PLD by software. This
sack of bits can be reverse-engineered so companies are not going
to supply these (you can extract those bits from a Win-Modem dll so
this gets a bit too ridiculous for some devices).

In most cases it's not some algorithm that needs to be protected.
Instead its the 400-or-so engineering man-hours used to develop the
contents of the PLD (notice I did not use the words "Gate-Arrays"
until now. There are many kinds of Programmable Logic Devices).

Until vendors stop being penny-wise-pound-foolish, we will continue
to have these kinds of problems. Vendor education will ultimately
be the fix, I predict.

Cheers,
Dick Johnson
Penguin : Linux version 2.6.10 on an i686 machine (5537.79 BogoMips).
  Notice : All mail here is now cached for review by Dictator Bush.
                  98.36% of all statistics are fiction.

On Wed, Jan 05, 2005 at 03:48:46PM -0500, linux-os wrote:

Well its mostly about making boards just a bit too cheap.

If the vendors would just put in a serial EPROM that loads the
proprietary stuff to their PLD upon startup, then there is
no "proprietary" code to worry about. You have the generic
published interface like a PLX chip, plus the specific published
register functions of their PLD chip. How the device actually
makes smoke and mirrors is hidden even from the programmer.

But, by eliminating the US$0.50 cost of a EPROM, they want to supply
a sack of bits that needs to be uploaded to the PLD by software. This
sack of bits can be reverse-engineered so companies are not going
to supply these (you can extract those bits from a Win-Modem dll so
this gets a bit too ridiculous for some devices).

In most cases it's not some algorithm that needs to be protected.
Instead its the 400-or-so engineering man-hours used to develop the
contents of the PLD (notice I did not use the words "Gate-Arrays"
until now. There are many kinds of Programmable Logic Devices).

Until vendors stop being penny-wise-pound-foolish, we will continue
to have these kinds of problems. Vendor education will ultimately
be the fix, I predict.

Well unfortunately I think you underestimate the cost of the serial
eprom needed to do the load on an fpga or pld or the like.  According to
our suppliers they are actually rather hard to get for some reason (that
I don't understand) and are more than $0.50.  The product I am current
working on software for would be a whole lot simpler if we could just
program an eprom with the fpga code, but instead we have to load it via
jtag through the parallel port to save a fair bit of moeny on eprom
chips assuming we could even get them in the quantities we would need.

The winmodem is just a sound card with the cpu doing the actual
algorihtm.  That is different and would not be solved with an eprom,
only with a real DSP or other chip capable of doing the modulation and
demodulation of the signals.  I guess calling them modem's is being too
generous, more like phone line sampler cards.

Len Sorensen

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wed, 5 Jan 2005, Luis R. Rodriguez wrote:

What I think we probably will have to do is just work torwards seeing if
we can come up with our own open wireless hardware. I know there was
a recent thread on lkml about an open video card -- anyone know where
that ended up?

This may be a silly point, but there *was* good 802.11g hardware available 
which worked well with the fully open drivers.  I presume the 
manufacturers are moving to the "softmac" design instead because (for 
them) it is cheaper.  However, the point is that the working designs are 
already there and it may be that buying the existing design which is being 
phased out is cheaper for the FOSS community than developing a whole new 
open device.

Maybe it would be possible to convince one of the manufacturers that it's 
worth their while producing the older design hardware - if there is a 
single manufacturer who is making more or less the only hardware that is 
guaranteed to work under Linux there is probably quite a market for them.

  - Steve       Jabber: steve@nexusuk.org     Web: http://www.nexusuk.org/

      Servatis a periculum, servatis a maleficum - Whisper, Evanescence

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (GNU/Linux)
Comment: Public key available at http://www.nexusuk.org/pubkey.txt

iD8DBQFB3Ex15zUOsIV3bqERAuInAKCGVS1kzaR4En2nQnKhDPv6TptZ+QCdEzFN
y8HbDEpnxvJql8AVpDePcnA=
=NfEa
-----END PGP SIGNATURE-----

On Wed, Jan 05, 2005 at 08:22:10PM +0000, Steve Hill wrote:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wed, 5 Jan 2005, Luis R. Rodriguez wrote:

quoted

What I think we probably will have to do is just work torwards seeing if
we can come up with our own open wireless hardware. I know there was
a recent thread on lkml about an open video card -- anyone know where
that ended up?

This may be a silly point, but there *was* good 802.11g hardware available 
which worked well with the fully open drivers. 

Yes, that would be the Full MAC prism chipsets with the linux prism54 driver,
which I help maintain.

I presume the 
manufacturers are moving to the "softmac" design instead because (for 
them) it is cheaper.

That is correct. They have already moved to the Softmac design and
you're lucky if you can buy FullMAC chipsets in stores now.

However, the point is that the working designs are 
already there and it may be that buying the existing design which is being 
phased out is cheaper for the FOSS community than developing a whole new 
open device.

Definitely, I agree. Anyone have an idea of how much buying a wireless
chipset design may cost?

Maybe it would be possible to convince one of the manufacturers that it's 
worth their while producing the older design hardware - if there is a 
single manufacturer who is making more or less the only hardware that is 
guaranteed to work under Linux there is probably quite a market for them.

I think they made the move because of economics as you mentioned
earlier. Under the current circumstances, I find it hard to be able to
Convince Conexant, for example, to start selling FullMAC chipsets again.

AFAICT the FullMAC chipsets have reached the END OF LIFE period.

	Luis

-- 
GnuPG Key fingerprint = 113F B290 C6D2 0251 4D84  A34A 6ADD 4937 E20A 525E

On Wednesday 05 January 2005 3:22 pm, Steve Hill wrote:

On Wed, 5 Jan 2005, Luis R. Rodriguez wrote:

quoted

What I think we probably will have to do is just work torwards seeing if
we can come up with our own open wireless hardware. I know there was
a recent thread on lkml about an open video card -- anyone know where
that ended up?

This may be a silly point, but there *was* good 802.11g hardware available
which worked well with the fully open drivers.  I presume the
manufacturers are moving to the "softmac" design instead because (for
them) it is cheaper.  However, the point is that the working designs are
already there and it may be that buying the existing design which is being
phased out is cheaper for the FOSS community than developing a whole new
open device.

Softmac doesn't necessary mean evil all the time. The adm8211 has a fully GPL 
driver for Linux. There is no firmware to worry about either. ADMtek actually 
helped me with the driver, by providing the source to their old 2.4 binary 
Linux driver and sending me (from Taiwan to New Jersey, fedexed overnight) 
two sample cards to test with. Oh yes, and it's quite cheap. (but 802.11b, 
which maybe a problem for some people)

If ADMtek can do it, why can't other manufacturers?

-Michael Wu