A career that started with software RAID
Dan Williams' involvement with the Linux kernel began with an internship, a copy of the source tree, and a personal dream of running RAID storage at home. The discovery that Linux had gained software RAID support was the spark. The ability to download it, use it, modify it, and contribute back was the hook. His earliest traces of activity in the kernel community go back to 2004, and from there the path forward was clear: professional kernel developer.
His early work focused on storage drivers, then expanded into something far more ambitious. Hardware vendors came to the kernel community with persistent memory, a category of device that didn't really exist yet, and asked the impossible: support it in mainline before silicon shipped. Williams took the assignment. Maintainers were reluctant to bend filesystems around a vendor moonshot. He bent them anyway, introducing DAX (direct access) as a new way for filesystems to talk to byte-addressable media and helping design a new filesystem optimized around the model.
CXL, trusted computing, and the next wave
Building on the persistent memory work, Williams led the introduction of Compute Express Link (CXL) into the kernel. He contributed to the specification and the concepts behind secure devices in the trusted computing context, areas that are now central to how hyperscalers think about memory disaggregation and confidential compute.
These are major technical achievements. The technical work, though, is only half of what the Linux Foundation's Technical Advisory Board is writing about. The other half is the human side.
The human side of the kernel
Williams served as both member and chairman of the Linux Foundation Technical Advisory Board. He contributed substantially to the interpretation of the Code of Conduct, the Kernel Contribution Maturity model, and the documentation of Inclusive Terminology. To anyone who has watched kernel mailing list threads spiral into something ugly, this is not abstract. The standards he helped shape are the reason a lot of people can still show up and do the work.
A former manager summed it up: he was the kindest and, at the same time, the most unwavering person on the planet. That combination is rare in a community where a lot of people are unwavering but not necessarily kind about it.
The person outside the code
Williams was a competitive Latin ballroom dancer, a passion that introduced him to Catherine, his wife. He and Catherine had three kids. The family shared a sharp, dry sense of humor: when Williams once found himself on a flight with Linus Torvalds, he texted Catherine to ask what she wanted him to say. Her reply was a flat "Merge windows suck!"
The memorial closes with Antoine de Saint-Exupéry: in one of the stars I shall be living. In one of them I shall be laughing. The Linux Foundation has opened a condolence address at djbw@tglx.de for anyone who wants to reach the family.
Kernel contributors tend to leave behind code, commits, and a few choice arguments in archived mailing lists. Williams left behind a community that, by every account, treats each other a little better than it did before he showed up. That is a larger legacy than any single subsystem.