Hi. I'm Dave Taht. I've been working for the last 28 months straight, nearly full time, mostly for free, at fixing the Internet's bufferbloat epidemic. I play Sancho to Jim Gettys' Don Quixote, doing whatever is necessary to keep things going.
I keep the bufferbloat.net website, servers, build cluster and mailing lists running, handle PR, co-ordinate the efforts of engineers and researchers across the globe, do research of my own, give talks, keep up with the literature, have found, fixed or help fix tons of bugs, and have written a bunch of code.
My biggest sub-project in support of the effort has been testing the feasibility of new de-bufferbloating ideas in a typical home router, with a bleeding edge version of OpenWrt, called CeroWrt. The R&D strategy (which is working!) is to get the ideas that work from cero into the mainstream and ultimately into a form that the typical home router vendor can adopt and ship.
If you are on this page, and don't know what bufferbloat is already, please google for it and come back. And this is a review of what CeroWrt is about, and here is Jim's best articulation of the Internet we're aiming for. Keeping this going eats me alive and that's why I have put up this page.
CeroWrt's goal is merely to prove that the various bloat-reduction technologies can, indeed, be deployed, across the internet, on devices big and small, in the places where it matters most - home gateways, head ends, wifi and 3g - and give those technologies away so that every manufacturer can use them, and every user of the Internet, benefit.
When I wrote this page originally (sept, 2012), I figured we have 3-6 months left before we had something truly great. And I was tapped out, dead broke, and ready to give up.
So I put out this plea for donations, and a lot has happened since then. We pushed out versions of the critical code into the Linux kernel and it is now the default in OpenWrt. Comcast Research picked me up for a few months and basically made me whole again, while we worked on things like the very detailed cablelabs study that looked in detail as to how to fix cablemodems, and CMTSes.
I put out multiple new releases of cerowrt, notably the "modena" release, which is still the current stable. (It was *almost* truly great. We've had some stability problems with getting the next release "berlin" out - which has a few things like a vastly simplified gui-driven AQM system) In addition to keeping a roof over my head in last september, the donations here have paid for about half the hardware now in yurtlab, which I've been blogging about periodically. My principle intent in building up the lab is to look hard at wifi finally, but a horde of other problems to solve are between here and that day.
Everything that works has gone upstream. Nearly every major bleeding edge Linux distro has fq_codel available as an option now.
Lastly, and most importantly, on July 30, 2013, the ietf voted in favor of forming a working group to analyze and standardize bufferbloat solutions. There were something like 180 people in the room, humming assent. It certainly is my hope that we will gain industry-wide momentum from that...
Yet there is still no "business model" here, the home router vendors are asleep at the switch, and still with or without financial aid to the effort, we're not going to withhold anything - we just want a better Internet to happen as fast as possible.
It certainly would be wonderful if the CeroWrt based research effort could become self sustaining, money dropped from the sky, and we could re-expand our scope to include everything we'd planned to do, but I'm not banking on it.
If you can help out, even a little, a financial boost goes a long way.
The above links are to my personal company (teklibre). It's a for-profit (not that it's made any). So cash sent this way is not tax deductable as a charitable contribution. From a goods and services perspective, perhaps I'm inventing a new category: "research as a service" - or "new firmware as a digital good". I'm trying to get out a quality release of CeroWrt every quarter (with numerous development snapshots inbetween), but the expansion of our outlook into more than just one router has ate into the time needed to get one router moving forward faster, and development of some new test tools, etc, has really slowed things down. I don't know what to do about it - to truly go and fix the internet - all 16 billion machines on it - is going to require a much larger team and funding than what we've got so far. ... but in the interim, your donations have helped keep the effort going, and top ramen on the table. Thank you with all my heart.
If you are looking for a charitable 501c(3) org to donate to instead, the Internet Systems Consortium was the earliest and most wonderful supporter of our quest to fix the bufferbloat problem. They donated 7 servers, rackspace, hosting, electricity, office AND lab space to our efforts. Without their ongoing support, the bufferbloat effort would struggle much harder than it has. ISC also supports some criticial internet infrastructure and a multitude of other open source projects. To contribute to their general fund, click here. If you are an organisation concerned with the future and ongoing health of the internet, ISC's got your back, and you can inquire as to how to become a friend of isc.
I (we) owe them a debt that I can never repay, although I would, if I could.
It is not (presently!) backed by some uber-rich zillionaire from his tropic island in the pacific.
Back in Jan, 2011, when Jim and I started it, we'd envisioned a 2mil/yr budget and a 6 person team with some of the greats in the industry onboard to tackle the bufferbloat epidemic. We then got too busy working on the technical side to form a 501c(3) for it...
Result: Total actual budget: 0. And a global effort consisting of hundreds of volunteers, contributing ideas, time, hardware, and code, to whom I try to give full credit in CeroWrt's credits page. We've made serious progress. Never has so much been accomplished so quickly on so little, by so many!
Frankly I couldn't have imagined a better outcome, technically.
There are other hard, capital/resource intensive projects, that we would like to be doing, that cannot easily be done by our extended community, so longer term, Bufferbloat.net's biggest "plan" is trying to structure one or more of the bigger, stalled out projects (network and wireless modeling, e2e measurements, testbed, field testing) into a format that could be put on kickstarter or indigogo, and to keep seeking the sizeable funding that we really need to kick those efforts into higher gear.
If you have any other long term funding ideas for bufferbloat.net's efforts, please share them!
In addition to my time, I'd spent a ton of money (when I still had some) on bufferbloat.net and CeroWrt. Karmically, given the positive results so far, I'm set for life. Financially, not so much.
If you've found my efforts on behalf of all this worthwhile, I could use some help making rent and getting new bits of needed hardware (a high end spectrum analyzer is the biggest ticket outstanding item) I would prefer to keep hammering away at fixing problems in codel and CeroWrt than do anything else, and to keep turning theory into something that can be deployed throughout the internet.
The bandwidth you save may be your own.