Keynote: Lightning Updates
summary
Claim: use content-addressable storage plus a graph of states and C++20/23 code to turn application-state updates into small, verifiable downloads. Implementation: strong-typed hashes, metadata/delta/snapshot objects, URL-able identifiers, protobuf/json blobs, cryptographic signatures, coroutine-based fetch, and atomic shared_ptr subjects for zero-dangling readers. At Avast scale—tens of millions of endpoints updated every few minutes—the server only needs to publish tiny hashes and clients compute/download deltas or snapshots; result is fast, auditable distribution (used for AV signatures and model distribution) with transitive trust via signed metadata.
AI-written summary · about
chapters (11)
- 0:00Start
- 11:44strong typing (proxy)
- 12:10strong typing.(with inheritance)
- 15:54metadata type
- 18:12identifier of an object
- 30:12selecting next step from a metadata node
- 36:31fetch path
- 40:55update_state
- 45:20update traits
- 47:48user API
- 50:51how is all the code used by a user?
related talks
11mKeynote: The Meteoric Rise of Sigstore
01h 19mKeynote: I Can Write the Code But Getting Something Done Is Another Matter
45mKeynote: State of the Cephalopod
33mApps and their Stores: An Alternative History
36mThe Power of the Paradigm
11mSoftware, but not as we know it: Part Two
41mKeynote: The Silver Bullet Syndrome Director's Cut - Complexity Strikes Back
59mKeynote
