[ath9k-devel] What mode to use for wireless backhaul?

[Felix Fietkau <nbd@openwrt.org>]

On 2011-04-27 2:18 AM, Peter Stuge wrote:
> Larry Vaden wrote:
>>  >  I was refering exclusively to the ath9k code and community.
> ..
>>  I look at ath9k, Felix, Adrian et al as real sw folks who have
>>  forgotten more than I'll ever know wrt the specifics of the matter
>>  at hand.
>
> Yes Felix in particular does an amazing job on the Linux driver, but
> the code seems to need *many* more years of work before it can be
> said to actually be finished,
Depends on what you mean by 'finished'. For many people it's quite 
usable and stable. The number of people that are still able to reproduce 
problems with ath9k on OpenWrt has been consistently decreasing over 
time, whereas more and more people are starting to use it, sometimes 
even in very big setups with thousands of users.

> I think likely it will not happen at
> all before hardware is obsolete. I was quite excited because I had
> the impression that Atheros was eager with Linux, but I expected
> something completely different from what I found to be the reality
> for ath9k code and Atheros' operation. I guess they do what they can,
> I just expected something vastly different.
>
> I'm eager to test ath9k hardware with fbsd because Adrian indeed has
> much more than a clue.
Sure, Adrian is good at what he does, but Atheros support in FreeBSD is 
still quite a bit behind, because for a long time very little work has 
been put into it - so don't expect too much, and if it doesn't work for 
you, please don't start whining on the FreeBSD lists as well :)

> The fact remains that ath9k on Linux is not mature rock solid in all
> different modes and bands, and that hasn't really changed over the
> last year and a half I've been around, despite the endless hours that
> have been spent on the driver. With regressions happening daily I
> have absolutely no expectation for Linux ath9k excellence in the
> future.
I get the impression that you're extrapolating from a very, *very* 
limited set of data, looking only at your own issues and the limited 
traffic on this mailing list, which only represents few actual users.

I'm looking at lots of feedback from a very diverse set of users, using 
ath9k in access point, ad-hoc mesh, point to point link setups, and some 
also using it in client mode.
Sure, not all problems have been resolved yet, but there's a large 
number of users for which the current code is completely solid and 
functional.

> I believe the code was simply too bad to begin with for it to ever
> work properly. Too many cases that someone never thought of when
> writing. I'd love to be proven wrong of course! I think the effort
> will total monumental however, which is such a waste of life. :(
Actually, a lot of the code quality issues have been resolved already 
and aside from a few things here and there (and of course the tx 
aggregation nastiness), the code is in pretty good shape.

- Felix