Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Rust, In Sickness & In Health

Avatar for Bryan Cantrill Bryan Cantrill
September 10, 2026
110

Rust, In Sickness & In Health

Keynote given at RustConf 2026 on September 10th, 2026 in Montreal. Video to come!

Avatar for Bryan Cantrill

Bryan Cantrill

September 10, 2026

Transcript

  1. 2017: Getting out of a bad relationship… Platform as a

    Reflection of Values, Node Summit 2017 OXIDE
  2. 2018: Lots to love! • Rust’s error handling is beautiful

    • I love tuples • The macros are incredible • The integrated testing is terrific • format! is a pleasure • The community is amazing • include_str! is a godsend • The performance rips • Serde is stunningly good OXIDE
  3. 2018: Dreaming of the future • In 2018, I knew

    two things: I wanted to one day start a company with Steve Tuck – and whatever we did would be in Rust • Importantly: we did not (yet!) know what we would do! • …but the values of Rust – robustness, performance, expressiveness – clearly resonated with the software that I wanted to build • It is not an exaggeration to say that we needed a problem befitting the language – and the technologists that the language attracted… OXIDE
  4. 2018: Rewrite the operating system in Rust? Is It Time

    to Rewrite the Operating System in Rust?, QConSF 2018 OXIDE
  5. 2019: The attraction of Rust • Our Rust use was

    not by fiat – but we also wanted to be sure that Rust was thoroughly disqualified before using a different language at any layer • This disposition served as a cultural beacon, attracting like minds • Attracting technologists with a disposition towards Rust greatly strengthened that disposition – become a positive feedback loop • We quickly discovered that it was more versatile than we imagined – and Rust itself became the lens by which we made technology decisions OXIDE
  6. 2020: Learning to live together • For myriad reasons (viz.

    RFD 26 Host Operating System & Hypervisor), we wanted a host operating system that was illumos-derived • This necessitated illumos Tier 2 support for Rust, which was… painful • Technically, this was straightforward – the challenge was getting fixes into the crate ecosystem upon which Rust tooling (e.g., rustup) depends • But it was also revealing: the Rust community was much more hospitable to alternative operating systems than other communities we’ve been in! OXIDE
  7. 2020: After the honeymoon • no_std • paste! • {:#x?}

    • unsafe • Integer literal syntax • Multi-platform support • DWARF support • anyhow! + RUST_BACKTRACE • Gimli and Goblin • asm! • Data-bearing enums • String continuations • Ternary operations • --pretty=expanded and cargo expand OXIDE
  8. 2021: Nest building Ben Naecker and Adam Leventhal, USDT Past,

    Present, and Future, dtrace.conf(24) OXIDE
  9. 2023: The brood • Embedded OS (Hubris) • REST framework

    (Dropshot) • Embedded debugger (Humility) • OpenAPI client gen (Progenitor) • Bootloader (phbl) • Storage service (Crucible) • In-kernel packet engine (OPTE) • Switch OS (Dendrite) • Hypervisor (Propolis) • Routing stack (Maghemite) • Control plane (Omicron) • P4 Compiler (x4c) OXIDE
  10. 2024: Reminding ourselves why we fell in love Adam Leventhal,

    Uniting Rust Servers and Clients through OpenAPI, Nordic APIs 2024 OXIDE
  11. 2025: A brief sickness Dave Pacheco and Eliza Weisman, When

    Async Attacks!, Oxide and Friends, 2025 OXIDE
  12. In sickness and in health • Broadly, Rust has been

    very good for Oxide: not only as an artifact and as a community – but as a reflection and projection of values • The grand Rust bargain – the shift of cognitive load from production to development – has been emphatically and unequivocally the right one • Async cancellation was a point of despair exactly because Rust has generally saved us from problems that require non-local reasoning! • Our most vexing technical challenges with Rust rhyme with one another: they stem from emergent complexity in production OXIDE
  13. Growing old together • It is very important to us

    that – above all else! – Rust not screw it up • We worry where Rust’s grand bargain may be undermined – and putative reduction in developer load exchanged for emergent production behavior • Debuggability – especially postmortem – can hold this in check (revisiting Kernighan's Law: if we can’t debug it, we shouldn’t write it!) • We are grateful for Rust and its community – they have allowed us to become who we are, attracting kindred spirits and like minds • We very much look forward to growing old together! OXIDE