By Fjodor on Apr. 11, 2007.
Late last year the technology blog The Jem Report attempted to shed some light on the state of wireless network connectivity under *BSD and Linux. In an article, author Jem Matzan addressed the problems for open source driver developers with regards to getting support from hardware vendors. He presented various drivers for Linux and *BSD and offered statements from driver developers and hardware manufacturers in the cases where he was able to get responses. Matzan stated that he had been unable to elicit any response from the MadWifi project, and speculated that this might be due to pressure from Atheros. MadWifi developers dispute that assertion, and claim they never heard from Matzan.
Getting wireless networking going under Linux can be a breeze or a hassle, depending on the network card or device you’re using. Complicating matters is the fact that, especially for wireless network devices, manufacturers seem either unwilling or legally unable to provide even the most basic information for open source developers as to how the hardware should be operated by drivers.
In the case of network cards using chips from Atheros, the people of the MadWifi project strive to deliver Linux drivers, and by most measures are quite successful at that. This author’s wireless home network is backed by a Linux box that provides Internet access through a MadWifi-driven Atheros card, and it works flawlessly. However, the MadWifi driver relies on a binary-only, closed source hardware abstraction layer (HAL) file from Atheros.
I spoke with Michael Renzmann, Will Herrick, and Kel Modderman, three members of the MadWifi team.
Fjodor: How would you describe you relationship with the MadWifi project?
Michael Renzmann: When Sam Leffler decided to abandon the Linux port of his driver back in 2005, I was one of the guys who stepped up to keep the project alive. Today my main task is to maintain the server that hosts various madwifi.org services, and I’m involved in strategic plannings related to the project.
Will Herrick: I do a bit of work on the wiki and help folks on #madwifi.
Fjodor: Would you have been able to answer Jem Matzan for the project?
WH: The questions regarding the MadWifi driver have had answers on public offer since well before the that article came out. Google or a browse of the Web site will satisfy all who wish to check that.
Fjodor: Jem Matzan claims to have sought to contact you. Did any such attempt actually get through to you?
MR: No, otherwise I would have answered.
WH: I saw no sign of him. No one did, as far as I can tell. We maintain an IRC channel, four mailing lists (with two online archives), a wiki, a Trac site, an SVN site, a Web page of contact and developer info, and generally have a visible presence on the Web.
Fjodor: Which modes of contact do you usually respond to? E-mail? IRC? The avenues described on the project’s homepage?
WH: All of them. You can see that by inspecting the archives. They are listed as resources and there is a link to that page on the wiki’s homepage.
MR: Even if contacting [us] directly failed, it would have been quite easy to get in contact with [us] through other channels, or to contact the project as such by sending an inquiry to one of our mailing lists. It did not happen in this case.
Fjodor: The article is mainly about company support for open source efforts to develop drivers. Have you had any contact with Atheros during your efforts?
WH: Not enough.
MR: Atheros has a history of supporting our work. One of the most notable events in that regard may be the contribution of a large amount of code that introduced support for multi SSID operation on one physical card (virtual access points). This contribution improved the functionality of MadWifi a lot and made it one of the most advanced WLAN drivers available for Linux today.
Fjodor: Have you sought contact with Jem Matzan in response to his article?
MR: Yes. I used his user forum to reply to the MadWifi-related part of the article. The whole thread got quite a bit of attention by the administrator (who, as far as I can judge from his comments, is Jem Matzan). However, he obviously didn’t care enough to answer my posting.
Fjodor: Matzan asserts little interest from you in supporting ar5k (the open source effort at making a HAL for Atheros chipsets). What are your views on ar5k?
WH: Once the ar5k base code passed its review for legal hygiene and was reported clean, it was only days before a new branch was added to the MadWifi SVN and work began on OpenHAL. Look at the commit log and you’ll see it’s been worked on steadily since inception.
MR: ar5k is used as base for OpenHAL, which is currently ported to Linux and integrated into MadWifi. Nick Kossifidis, who is the maintainer of this port, has joined the MadWifi team and now maintains his work directly in our Subversion repository. It’s our long-term goal to get MadWifi into the mainline kernel, and the results of Nick’s work on OpenHAL are definitely an important milestone for that goal. That was discussed at length around December 2006, for example.
I fail to see how Jem came to the conclusion that ar5k had “virtually no support” from the project.
Fjodor: Do you have any closing comments?
MR: I’m disappointed by the quality of the article, at least that of the part about MadWifi. There was a public response to his article, telling him where he was wrong and why. The fact that he didn’t bother to react to that, in my eyes, demonstrates a lack of interest and professionalism. That leaves me wondering about the reliability of the rest of the article.
Matzan’s allegations regarding pressure from Atheros to squelch the ar5k/OpenHAL’s inclusion in MadWifi are wrong. He should have easily spotted that while writing the article, even without getting in direct contact with us. To wrongfully cast Atheros (on /. no less) as a company discouraging FOSS wireless helps neither the Linux community nor the project, but only discourages this company from continuing and expanding their support.
All this is sad, because this would have been a great chance to explain the FOSS WLAN issues that face driver authors and chip makers every day. And in case Jem changes his mind, I’ll still be available for a public discussion of his questions.
WH: Atheros deserves recognition and credit for their participation in BSD and Linux wireless driver support. The current MadWifi code base is derived from source and documentation provided by an Atheros code drop, for example. The BSD work goes back several years. Rewarding Atheros for their past help might encourage them to take future FOSS steps.
The author contacted Matzan to give him a chance to comment on this article. The following reply to the initial mail came forth, and none on a subsequent request for a clarification on the nature of the more substantial assertions and the lack of a reply to Renzmann’s comments on the article:
Jem Matzan: The only thing I alleged [sic!] about the MadWifi developers was that they were difficult to contact. I phoned several times, left a voice mail message, emailed a few different addresses listed on the MadWifi developers page, and all I got was silence or mailer daemon bounces. That’s really all I’m qualified to comment on regarding the matter.
Disclaimer: I am not directly associated with the MadWifi development team. I am, however, a long time user of their driver, and a regular visitor to the #madwifi channel at irc.freenode.net.