Re: [Madwifi-devel] Re: Proper contact at Atheros...

[Mathieu L. <Mat...@so...>]

Hi greg,

comments interspeced below.

On Thu, 2003-09-25 at 20:18, greg chesson wrote:
> In order to work on the HAL, a developer needs the same information
> (and support) provided to  vendors who are building boards and
> systems and releated software.   It is not a task to be undertaken lightly.
> It is not a task that can be undertaken with a "no support provided" 
> agreement

I have already done similar work on other rather complex systems. I am
not scared by lack of support. I am scared by lack of control (not in
the RMS sense, in the development sense) of a source code.

Basically, if there is a bit of code I do not control in my driver, how
can I debug something in it ? I have stability problems with the current
atheros driver and I have the technical knowledge to fix it but I cannot
do so because binary code makes a driver a black box (I learned this
through personal hard experience).

> because  hal developers need access to the chip designers as well as access
> to lab test facilities - spectrum analyzers, calibration equipment and 
> other gear

I do not think so. I do not plan to rewrite the HAL. I just need to see
the code and be able to single step through it from time to time when
debugging my own code.

> Right now there are about a dozen companies with HAL source.
> They have full Atheros support, are building boards and systems,
> and generally know what they're doing although they still need assistance.

I find this very insulting. You are basically saying the the open source
community is incompetent technically.

> These companies will also be certifying their Linux-based products with 
> regulatory
> agencies.   That is not the case with the majority of folks who would like
> to hack on the driver.

I understand there are regulatory issues with making certain parts of
the HAL public. However, it seems to me that a random NDA saying that,
I, as a developper, will not distribute the sources or binaries for
these parts should be enough with regard to the FCC or any other
regulatory agency.

> 
> There is some hal porting work to other architectures that will appear
> on sourceforge.   That will satisfy most requests for hal source.

I am not interested in other architectures.

> Other improvements are coming along, mainly exposing more chip functionality
> for crypto, qos, and power control.   These will also appear on sourceforge.

I am not interested in depending on other people's schedules, even more
when these schedules are not publicly available. 

> 
> In short - we welcome competent help:
> obtain a technical license agreement from Atheros,
> show up with some patches that do something worthwhile (paraphrasing Linus),
> and we'll be grateful.

In short, you are saying, pay 40KUSD, write code for us and we will like
you. I find this very rude and rather unproductive.

I have immediate needs for further functionality in this driver. I did
not ask anyone to add this functionality to the driver because this is
Open Source: just hack it, send a patch and everyone will be happy. I
have the time (because I was assigned to this task at work) and
technical knowledge to implement these features. Because I am not a
super guru, I write buggy code all the time. This buggy code must be
debugged. Debugging my own buggy code without being able to see what a
given piece of code does exactly is a waste of time. 

Experience shows that the bug always comes weird interraction between
the black box and your code because either you are the first one to
really use the feature or the first one to use it that way. The
existance of this binary code/black box makes it impossible for me to
debug my own code.


To summarize:
  - if there is a schedule for other platforms, new functionality, make
it public.
  - please, make public as much source as possible from the hal.
  - please, document as much as possible the _exact_ behavior of each
function in the HAL and its _side effects_.

Greg, I am not yelling at you because you are evil and write proprietary
software, I am not yelling at you because I find your answer very rude.

I am simply stating that the existance of a driver with binary bits
makes it impossible for _developers_ to work on this driver. As such, I
wonder why Atheros has tried to make this driver free software. 

Is it because they hope to get support from the Open Source Community
(and thus get free -as in beer- external software developement
ressources) ? If so, this was a mistake I think. 

Is it because they need to provide end-user support for the cards based
on atheros products ? If this is the case, it would probably be way less
painful for Atheros to release binary-only drivers and not try to
release half-binary drivers. It is not less legal with regard to the
Linux kernel GPL and it would save you the pain of listening to weenies
like me.

regards,
Mathieu
-- 
Mathieu Lacage <mat...@so...>