make-wifi-fast: would anyone at canonical want to help out?

Dave Taht dave.taht at
Mon Mar 9 22:09:04 UTC 2015

Since most of the core ideas and theory we have for vastly improving
wifi have landed (at least in theory), we are in the process of
kicking off the make-wifi-fast project to get the fixes done, and
tested, and "out there"...

First up was minstrel-blues (do check out the patches!), then there
are some tcp related fixes in progress from google, and after that is
per station/vif queuing and a ton of stuff to reduce buffering and
more sanely use up txops described on pp 25- here:

Would anyone at canonical be interested in participating in getting
all this new stuff developed, tested, and "out there"?

There is an awful lot to it. I had described most of the basic
problems in my talk at MIT 2 years back:

and now I think the fruits of the effort can now not
only be applied to isochronous ethernet-like devices with fq_codel
everywhere, but also to wifi (which is very different, and has
required some rather fundamental changes to the queue management
algorithms, that are going to take a while to ingest and debug)

We can make wifi a lot better by addressing a ton of low-lying fruit
in the stack. We think. With a little more help. Anyone interested?

More information about the kernel-team mailing list