[Cerowrt-devel] Bufferbloat and other Internet challenges - was Congratulations

Dave Taht dave.taht at gmail.com
Tue Sep 16 01:31:48 EDT 2014

On Wed, Aug 20, 2014 at 12:09 PM,  <dpreed at reed.com> wrote:
> I just read Vint Cerf's latest column in IEEE Internet Computing (October
> 2014), paper edition that arrived in the mail.  Entitled "Bufferbloat and
> other Internet challenges", it highlights Jim Gettys, Dave Taht, Eric
> Dumazet, Codel/FQ_code, the CeroWRT/OpenWRT team, etc. pointing out that

That column is now publicly available.


I appreciate the honor he's done us, and also want to publicly
appreciate the trailblazing he and so many others did on the Internet
itself. Getting IPv6 to work as well as we have in cerowrt and
openwrt, was sort of a way to say thanks for all that...

> "The use of open source software to promote broad adoption and use of new
> technology is now well demonstrated ... The CeroWRT/OpenWRT effort could
> have a similar effect, especially if the resulting software can be ported to
> a variety of hardware platforms."

Openwrt is being built, daily for 36 platforms, including most common cpu types,
and hundreds of different chipsets, and 1000s of products can be
upgraded to use it.


(This build farm (mostly hosted in the donated google cloud)
occasionally breaks, and could use a little maintenance as I write.
The barrier breaker release is being done on another cloud entirely,
donated by another party )

This is still a small drop in the bucket, compared to the market today, and only
rarely includes the most bleeding edge of chipsets as only a few of
which use openwrt as a base.  But a lot of the heavy architectural
lifting is already done, and certainly much landed in the mainline
kernel and user space applications that will land in other Linux

We have billions of machines to upgrade, eventually.

> This is a quote and a column you can "take to the bank", especially the "if
> ... can be ported to other hardware platforms", which it should be.

I'm not sure what "bank" would offer credit based on a project as
speculative as this one has been. When we started we didn't know
if we could make ipv6 deployable on CPE, or how to fix bufferbloat,
nor if crowdsourcing the problems would work, or that the open source
community as a whole would "get it", or that various companies could
or would contribute resources, etc.

With an end goal is something as diffuse as "saving everybody on the
internet a few seconds a day of time, annoyance and frustration", the
multiplier effect translates out to 10s of thousands of man-years,
saved, eventually,
depending on how you do the math. But somehow these sort of
incremental improvements have to be demonstrated and then moved into
products, and it's hard with a quarterly vision, to think and work
ahead of the curve that far.

We've treated the entire world as a PARC.

I am reminded of Jay Leonhart's song about Robert Frost:
 ( http://www.jayleonhart.info/images/Robert_Frost.pdf )

"How did Robert Frost make payments on that country house of his?
 Where did he get the dough?
 Could he go down to the country store and sell a poem
 'saying, here's a nice one I wrote about the snow'..."

Maybe going out, hat in hand, for the next projects (like make-wifi-fast)
will be easier, now.

> Congratulations, all.

The only currency I've ever had to pay anyone with, was credit, and
I've done my best to credit everyone I could, here:

Thank you, all. If I missed anybody in that list, let me know. The
updated credits file is on the router software itself.

> I have said before that this project's technical
> contribution impresses me as much as Bram Cohen's creation of the original
> BitTorrent, and that's no small comment.  I nominated him for a number of

/me wipes away small tear. Thanks for playing with us, too, David.

> awards at the time, and he got at least one (he would have gotten "best
> paper of the year" if he had published his paper in an ACM journal, in my
> opinion.)

Some of the core papers driving this work were
unpublishable in today's academic environment, for the record some were:

Van and Kathie's: "RED in a different Light"
Paul McKenney's extended SFQ paper
Andrew Mcgregor's original work on the minstrel wifi rate controller,
- still unpublished
Toke's work on evaluating the various algorithms with netperf-wrapper
- the first leaked, the second, more comprehensive one, was recently rejected.

Are there awards for papers that couldn't get published that nonetheless
influenced the future of the internet? Is there an archive of those?

> Now Vint's opinion is supporting your work as well (and that has
> real weight).

I think "a" next step would be to try and write up what worked,
and what didn't, in the hope that the methods that did could be
embedded into more research, development, and engineering

There is still a great deal of work left to do to "cross the chasm",
towards steady, incremental gains in internet performance,
in a post-bufferbloated era. And certainly Vint's second point,
that embedded devices be made more easily upgradable, is
yet to be addressed in most of the Internet of Things.

Dave Täht


More information about the Cerowrt-devel mailing list