![]() |
“Elon, let me inside this thing, and I can fix it for you” (Image
from this podcast) |
Dave Taht died on April 1st. I met him only recently, and
never in person, but his passing saddens me. His technical work and
evangelism have improved the Internet, and I will give some examples of his
contributions to the Internet community and users, but I am sad because he was
a good person — idealistic, unselfish, open, and funny. I'll miss him. First, his contributions, then his values.
Contributions
Taht was best known for his work on buffer bloat and its impact on Internet performance. As packets hop across the Internet, they are queued in buffers while waiting to be forwarded. Long queues, “buffer bloat,” means increased latency, transit time between a source and its destination, and jitter, transit time variance.
Internet service providers typically advertise and price based on service speed, but latency is critical to interactive applications like gaming, teleconferencing, and Web surfing. (For a deep dive on buffer bloat, see this post by Jim Gettys, who coined the term buffer bloat.)
With Jim Gettys, Taht co-founded the buffer bloat Project, where he implemented, tested, and integrated active queue management (AQM) algorithms CoDel, FQ-CoDel, and CAKE, and led the CeroWRT (Customer Edge Router Wireless Router) project that focused on home, office, and other edge networks.
Gettys shared Taht's focus on the edge, writing “Surprising to most, AQM is essential for broad band service, home routers, and even operating systems: it isn't just for big Internet routers” in 2011 and more recently he said “Buffer bloat can happen anywhere in a network, though by far the most common locations are before/after the WiFi hop in the router, and then the hop from the home router back to the ISP”
Given his interests in space and edge networks, Starlink was a natural focus for Taht, and he talked about Starlink in this 8-minute podcast excerpt. Taht had known Elon Musk since he had worked on an ill-fated satellite that was on a Falcon 1 rocket when it blew up, so he emailed Musk in 2013 and offered to help, but Musk was not interested. The plan for Starlink was announced in January 2015, and when it eventually entered beta in 2020, Taht learned of the latency problem and emailed Vint Cerf, who arranged a meeting with Starlink engineers on Taht’s boat, but he did no work for them.
In January 2024, Elon Musk announced that “the biggest single goal for Starlink from a technical standpoint is to get the mean latency below 20 ms." By March, they were delivering results and listing latency as well as upload and download speed on their availability map. Unfortunately, Musk and his engineers did not listen to Taht in 2013 and 2020.
![]() |
Starlink latency improvement after January 2024 (Source) |
Values
The National Science Foundation backbone network (NSFNET) was created in 1986
to serve research and education in the US. The network grew rapidly, and in 1988, access was expanded to include
international research and education organizations. By 1996, 28 nations were
connected. (Fun fact – Cuba’s
first link was to NSFNET.)
Taht was 30 years old when NSFNET was decommissioned and the
Internet transitioned to fully commercial operation, but his values were
established by then. With his skill and experience, he could have found
lucrative work or built a large company, but he was committed to open-source
software and universal connectivity.
Everything he wrote was open-source, including his songs. He
is surely the only composer/musician to write geeky songs about the Internet, for example, this little ditty about the GNU Public License, Richard Stallman, Linus Torvalds, and Eric Raymond. He also spent years in Nicaragua, trying to find ways to bring the Internet (and power, lighting, food, medicine, and books) there as an
outgrowth of Nicholas Negroponte’s One Laptop Per Child
project.
He understood J. C. R. Licklider’s conception of the Internet as a means
of creating communities of common interest rather than common location. He created a Starlink list, which
is where I met him. The caliber of
conversation on the list is an implicit tribute to Taht.
The photo of Taht at the top of this post was taken from the
podcast in which he offered to fix the Starlink router, which was in beta at the
time, for free (though he wouldn’t mind a thank-you tweet, a new motor for his
boat, or even a Christmas card). The photo illustrates his values. He could have
lived in a mansion, but he chose to live on a boat with a guitar nearby. To know Dave Taht better, watch
the entire video.
Appendix
I met Dave recently and only knew him online, so I asked Perpleity, Gemini, Grok, and ChatGPT to list memorials. They returned several broken links, and Claude and Deepspeak were not current. I found these:
Author: Doc SearlsTitle: Remembering Dave Taht
Link
Title: Dave Taht, Who Sped Up Networks More Than You’ll Ever Know, Has Died
link
Author: Tom Stricx
Title: Honoring Dave Täht and his contributions to a better Internet (video calls included)
Link
Authors: Robert, Herbert, and Frank LibreQoS
Title: In loving memory of Dave Taht
Link
Author: Hacker News Community
Title: Thread collecting memories, technical anecdotes, and condolences from the networking and open source community
Link
Author: Toke Høiland-Jørgensen
Title: Remembering Dave Taht, Link
Link