Re: [Discuss-gnuradio] Re: [dttsp-linux] Intel ATOM WHOOAAAAA Nellie

[Dan Halperin]

On Feb 19, 2009, at 9:44 PM, Gregory Maxwell wrote:
> Nvidia video is a pretty poor choice for Linux too— you're tied to
> their proprietary drivers which often cause weird bugs (usually the #1
> cause of kernel panics on the kernelopps data collection project,
> right ahead of a proprietary wifi driver), and that driver ties you to
> whatever kernel and xorg versions they are willing to support.

I just have to jump in here and say that you're being pretty dogmatic.  
I'm not disagreeing with you on the fact that Nvidia doesn't support  
Linux in an ideal way, but you should rethink what you mean by  
proprietary or perhaps by the connotation in which you use it.

What is a 'proprietary wifi driver'? Better yet, what isn't a  
proprietary wifi driver? In particular, (assuming you were digging at  
iwlwifi, which has been near the top of kerneloops recently), the only  
thing `proprietary' about the newer Intel WiFi drivers is a binary  
firmware component that, primarily, enforces the FCC regulations. **  
See Disclaimer [1] **

The FCC's attitude towards enforcement is exactly the same thing that  
is worrying the SDR community [2], but which we currently get around  
it simply by ignoring it [3]; we can do that since Ettus Research  
isn't (yet :) exactly a huge multinational corporation. Intel, on the  
other hand, does have to worry about such things.

Here's what I do know about Intel & the iwlwifi drivers:
- Intel's corporate position (fairly newly revised) regarding WiFi  
drivers is that the day Windows drivers are released, functional Linux  
drivers need to be released as well
- They're developed by a small (ridiculously small, IMO) team that's  
relatively new to the Linux kernel game, but they're learning more  
about the process and 'doing more things right' every day
- They've proved willing to modify drivers to adopt suggestions from  
OS community, (e.g., adding support for injection mode)
- They've got a very happy set of users due to their being very  
responsive on the Linux driver mailing list
- iwlwifi is (IMO) the most mature and feature complete 802.11n driver  
implementation of any of the major vendors
- And because iwlwifi has implemented most of the new 802.11n features  
first, they work closely in tandem with the main kernel wireless guys  
(e.g., mac80211 developers) in defining the major interfaces in the  
networking stack and adapting them for universal compatibility [4]

What more could you want? Your flippant remark above is exactly that  
kind of muleheaded, black-and-white attitude that will keep other  
large corporations like Intel from warming up to the open source  
community faster. And that's a real shame, because when these policy  
shifts like Intel's do happen the results are often pretty great.

Dan

[1] ** Disclaimer ** I interned at Intel Research Seattle this past  
year, and have conducted research using the iwlwifi driver and also  
contributed bugfixes back to the same. Everything I say here is  
completely my own opinion and none of it represents Intel Corporation.  
[But some of this puts me in a good place to inform you about their  
process and attitude towards OSS and the Linux community.]
[2] See e.g. this old GNU Radio thread: 
http://www.mail-archive.com/address@hidden/msg00559.html
[3] You can trivially bypass USRP FCC enforcement, and Matt will even  
tell you how :) e.g. http://www.mail-archive.com/address@hidden/msg14203.html
[4] See aggregation support, which was in the iwlwifi drivers before  
Atheros announced open source ath9k drivers. However, the kernel  
network stack wasn't easily able to support this and they're still  
working on fitting it in.