A peer-to-peer, multi-writer data store.

Specification of a hash function that allows for verified streaming, similar to BLAKE3’s Bao.

A prefix authentication scheme that is more efficient than the one used in certificate transparency. Supersedes Bamboo.

A macro system for generating plaintext files. Suitable for producing web-native scientific papers such as this one. And the website you are currently reading, of course.

A Rust library of abstractions for lazily producing or consuming sequences.

A tool for scheduling meetings across timezones, mapping UTC offsets to emoji. Carrying on some of @cinnamon’s light.

A family of randomized set data structures that generalizes the zip-trees to work more efficiently in the presence of cache hierarchies. G-trees have applications for set fingerprinting and membership proofs.

Range-based set reconciliation is a simple approach to efficiently computing the union of two sets over a network, based on recursively partitioning the sets and comparing fingerprints of the partitions to probabilistically detect whether a partition requires further work.

Range-based set reconciliation used to require non-standard hash functions with certain homomorphic properties. This paper describes how to do away with these and use regular Merkle trees instead.

APIs like iterators, streams, sinks, readers, or writers are all APIs for lazily working with potentially infinite sequences of items. There is a large number of competing designs, but little principled work on evaluating them. Starting from the question of how to objectively evaluate such APIs, I derive some instructive and useful hierarchies of interfaces for lazily working with sequences.

Like certificate transparency logs, but more simple and more efficient.