LWN: Comments on "Atheros hires ath5k developer"

Atheros hires ath5k developer

opheriper — 2008-05-21T20:52:26+00:00

congratulations, Luis!

I've got a job offer from Ahteros too, and another offer from Cisco, both are ASIC design
position. Now I'm confusing which company I should join, could you guys give me some advice,
thanks very much!

I wonder if it is market pressure

muwlgr — 2008-05-17T11:07:24+00:00

S*it, how wrong I was. Asus Z99Le uses A8LE motherboard with Atheros 5006EG (or 5007EG?) card
: PCI ven:dev 168c:001c, subsystem 1a3b:1026 . Never ever buy it in your life. There will be
neither working madwifi/ath5k driver, nor anything working opensource from Atheros. Trust me
:>

Atheros hires ath5k developer

MisterIO — 2008-04-22T14:04:24+00:00

I wasn't talking about that.

I wonder if it is market pressure

muwlgr — 2008-04-19T15:30:19+00:00

ASUS Z99 is a great serie. They are all-Intel models (chipset, video, wireless), they are
light, and they are sold without Windows. ASUS definitely has to compete with itself.

I wonder if it is market pressure

ofeeley — 2008-04-18T16:11:48+00:00

I know that when I read your posts about it I decided to hold off on getting one.  I purchased
another machine instead (which has an intel3945 wireless chipset and as a result works
wonderfully with Fedora 8 and Fedora 9 Preview's version of NetworkManager.) I think a lot of
other people that are interested in using GNU/Linux are at the stage where they realize that
it's easier to buy hardware that's supportable by the community AND that hardware actually
exists -- mainly from Intel.

I wonder if it is market pressure

drag — 2008-04-18T12:29:23+00:00

> What I am wondering about, is if this is happening because Atheros is losing business to
Intel on a scale that makes it worthwhile to hire a developer.

I know that with my EEE I ended up with a Atheros wifi. It wasn't even supported under
Madwifi, Xandros used these special 'swan' drivers I haven't heard of before.

So I know that lots of Linux people were upset about that and it required the use of
ndiswrapper for a while till Atheros released a special patch specifically to allow Madwifi
driverrs to work. I told other people before the EEE came out that I was dissapointed that
Asus choose Atheros over Ralink or Intel, of course they all thought Madwifi was cool... until
after the EEE came out the Madwifi folks were completely powerless to support the hardware
until that HAL hack was released by Atheros.

Keep in mind that a angy customer making a support call is enough for Asus to loose
profitability on that EEE sale...

So they defenately were paying attention and it was a big enough issue that they (Atheros) had
to do a abnormal release.

I expect that other manufacturers were paying close attention to the EEE and they would of
noticed the problems with having to support a single-kernel-specific proprietary module and
the problems it caused OEMs and customers.

Where as if they used Intel (or a handful of other manufacturer's cards) then it would of been
a complete non-issue. I expect that this has caused Atheros to loose out on a couple big
potential customers.

I wonder if it is market pressure

tarjei — 2008-04-18T07:47:24+00:00

This is good news. 

What I am wondering about, is if this is happening because Atheros is losing business to Intel
on a scale that makes it worthwhile to hire a developer. 

One developer may cost somewhere around $200 000. You do not have to loose a lot of sales
before that becomes the cheapest option. 

With big OEMs like Dell and IBM focusing a lot more on Linux support on desktop machines than
they used to combined with Intel providing a very good alternative we might be seeing a
turning point when it comes to support for wireless cards.

Broadcom

mb — 2008-04-17T16:26:38+00:00

> In my personal view, broadcom provide a more stable driver than atheros, 

Am I missing something? Where does broadcom provide _any_ driver? (aside from binary MIPS
drivers and such stuff for embedded devices).

Broadcom

mjdcl — 2008-04-17T15:47:39+00:00

In my personal view, broadcom provide a more stable driver than atheros, I got atheros src
code, there even has a bug on Makefile...  it can't work on my board, reset always.

Atheros hires ath5k developer

johoho — 2008-04-17T07:07:51+00:00

it was never a secret that showing a company of how good you understand their own products
puts you in a good hiring position.

Atheros hires ath5k developer

pabs — 2008-04-17T06:22:48+00:00

Found this link after googling for the crda thing:

http://lwn.net/Articles/271037/

Atheros hires ath5k developer

pabs — 2008-04-17T05:59:46+00:00

Isn't Atheros HAL/OpenHAL something that runs on the host CPU?

The folks at prism54.org are developing FreeMAC, free (GPLed with source code) firmware for
Conexant wireless chips. Seems they have it easy though, since the Prism54 wireless cards have
an ARM CPU on them.

So, as long as wireless devices don't require firmware images to be signed before running them
(like bitfrost), we'll likely be able to try to write free firmware - and violate the
regulations :)

Anyways, I don't see how one firmware image can know which country it is in and which
regulations to comply with, nor if the user has a special licence to do software
radio/hamradio or the like. Even a kernel driver probably wouldn't be able to know that
without some kind of configuration info passed to it.

Atheros hires ath5k developer

MisterIO — 2008-04-17T05:49:19+00:00

So, does that mean that reverse engineering is even more useful than what was supposed to be?

Atheros hires ath5k developer

muwlgr — 2008-04-17T05:09:58+00:00

You mean, like "Creative hires Daniel_K to stop him hurting their business strategy" ? :>

Atheros hires ath5k developer

ikm — 2008-04-17T03:05:32+00:00

My thoughts as well

Atheros hires ath5k developer

BenHutchings — 2008-04-17T01:46:20+00:00

"Intel are NOT that great, their wireless drivers still require binary-blob non-free
firmware."

So far as I know, all wireless Ethernet controllers require firmware. Some of them load it
from flash, but since no-one expects to boot over wireless Ethernet, that's an unnecessary
expense. The last controller I know of that booted from flash was the Ralink RT2560, which has
been superceded.

I don't whether the firmware fundamentally has to be non-free. Radio regulations in some
countries might require that. Even so, it will tend to be so interdependent with the hardware
as to be practically unalterable by outsiders.

Atheros hires ath5k developer

pabs — 2008-04-17T00:59:05+00:00

Intel are NOT that great, their wireless drivers still require binary-blob non-free firmware.
At least their binary blob doesn't run on the host system any more, but fixing this issue
isn't something we will see without any reverse engineering by people outside of Intel:

http://bughost.org/bugzilla/show_bug.cgi?id=1594

I'd dearly like to see something like ath5k's OpenHAL or prism54's FreeMAC for Intel wireless
cards, but I wouldn't have the foggiest idea about were to find information about the
hardware.

Atheros hires ath5k developer

okeydoke — 2008-04-16T22:20:50+00:00

The OpenBSD Atheros driver could use some work too. It doesn't support 802.11g or WPA.

Atheros hires ath5k developer

nowster — 2008-04-16T21:54:38+00:00

It appears to be tweakable via debugfs. Linux Wireless Wiki

Atheros hires ath5k developer

drag — 2008-04-16T20:43:22+00:00

> Hopefully Luis will be allowed to release his efforts in a form that can be incorporated
into the stock kernel source.


That's what the man said in the article, that Atheros is trying to get driver support into the
vanilla kernel.

Atheros hires ath5k developer

k3ninho — 2008-04-16T19:57:45+00:00

Ah.  An explanation for my rt2500pci mac80211-based driver falling back to low speeds and
requiring a 'iwconfig' at the command to get the device back to full speed.  Thank you!

Atheros hires ath5k developer

nowster — 2008-04-16T19:05:28+00:00

Hopefully Luis will be allowed to release his efforts in a form that can be incorporated into
the stock kernel source.

At the moment the semi-free madwifi is more reliable than ath5k with 2.6.25-rc kernels[*], due
(I suspect) to problems with the mac80211 layer being rather too aggressive in reducing the
link speed in face of less than ideal signal conditions. Signal quality/level monitoring
(using wavemon) suggests that unsmoothed values are causing the rate algorithms to believe
that conditions are worse than they actually are. I see similar problems with the in-kernel
rt2500 driver.

[*] Incidentally, the 2.6.25-rc8 ath5k driver doesn't install, due to symbol mismatches.
compat-wireless does work.

Broadcom

mb — 2008-04-16T18:42:08+00:00

> O Broadcom, where art thou?

Yeah, come on. I also need a job. :P

Broadcom

elanthis — 2008-04-16T18:29:45+00:00

O Broadcom, where art thou?

April 1

corbet — 2008-04-16T18:22:42+00:00

Sorry for the hedging there - for some reason, something about the message set off my hoax alarms, to the point that I held it for a while to see what followups came. I've removed the "April 1" reference from the blurb.

Atheros hires ath5k developer

JoeBuck — 2008-04-16T17:41:47+00:00

If a company releases even a buggy driver, it still helps, because it provides information that the reimplementers need to know.

Ideally, of course, a good driver would be produced by effective two-way feedback between the experts on the hardware, and Linux kernel experts. One way is to get the company producing the product to hire an experienced Linux kernel developer to bridge the gaps, so good for Atheros.

Atheros hires ath5k developer

BenHutchings — 2008-04-16T17:26:06+00:00

Ralink did release source for their Linux drivers, but that showed just how buggy they were.
The drivers that have made it into Linux upstream have been reimplemented by others.

Atheros hires ath5k developer

drag — 2008-04-16T17:16:04+00:00

Wow.

Go Atheros!

With them joining the ranks of companies like Ralink or Intel (sorry those are the ones that I
use, if there are others with good open drivers I want to know about it!) for providing good
open source drivers for Linux this is a very cool thing.

And special thanks for OBSD for providing the open source 'HAL' for Atheros.


Now this along with open source video driver support from Intel, AMD, and Via (hopefully..)
then things are looking up for these two traditionally sore spots on with open source drivers.

Amazing.

Atheros hires ath5k developer

mcgrof — 2008-04-16T17:10:47+00:00

Certainly not an April 1st joke :)