I don’t see myself making major contributions to (rust-) libp2p in the near future and thus I am stepping down as a maintainer.

As announced before, I have left Protocol Labs in December 2023. After a 2 month re-orientation break, I have decided to move on entirely.

My first commit was 5y ago, a small bug fix in our address handling. Since then lots happened. A couple of milestones I was involved in:

  • First major contribution was an implementation of S-Kademlia’s disjoint paths. While a good first project, it was never deployed on any of the larger networks. Attacking a DHT is not that hard, making it secure takes a lot more than this.
  • Eventually we added what one would expect every p2p library to have - support for hole punching. That was a huge effort. We rolled our own relay protocol, a custom coordination protocol, an address discovery protocol, … Long story short, we invented lots of new wheels. Really, we should have started off of ICE. E.g. instead of designing our custom address discovery protocol, we should have learned early on from STUN and its past mistakes. This stuff is hard.
  • rust-libp2p was long missing a QUIC transport implementation. Eventually we were able to deliver this based on the quinn project, which I am very happy about, bringing lots of performance improvements to rust-libp2p users.
  • I managed the specification and implementation of WebRTC in libp2p. While a great match for a p2p library, unfortunately up until today, it has not seen larger adoption.
  • A tiny fraction of IPFS today is powered by a rust-libp2p based IPFS bootstrap node handling ~35k connections on a medium-sized machine.
  • A major portion of my time in 2023 went into an automated performance setup for the many libp2p implementations. The outcome in my eyes is both visually appealing and impactful from a performance perspective (see final item).
  • We shipped many developer experience improvements, taming some of the complexity inherent to peer-to-peer networking e.g. the SwarmBuilder.
  • Security is hard, especially on the internet, doubly so when in a peer-to-peer setting. A lot of our work went into security fixes, especially around DOS attacks. See e.g. a major memory exhaustion attack and a talk inspired by this work.
  • Talking about talks, I gave several on peer-to-peer networking and libp2p as well as helped organize related conferences, e.g. the FOSDEM networking devroom.
  • As a final big contribution, the before mentioned automated performance benchmark setup enabled me to land a major improvement to libp2p’s bandwidth performance. Implementing stream receive window auto-tuning in the Rust Yamux implementation improved throughput on high bandwidth connections from 33 Mbit/s to 1.3 Gbit/s.

During my time the rust-libp2p maintainer team went from 4 down to 1, up to 2, 3, back to 4, and now down to 1. I hope for another positive trend to follow. I enjoyed working with every maintainer, the larger libp2p team and the many contributors. Thank you! None of the above listed work would have been possible without you. None of it I could have achieved on my own. In the greater picture of the libp2p project, this list is negligeable, but probably fun for me to look back at in a couple of years from now. (👋 hi future Max.)

For more stats:

I will stay subscribed to the repository for now to answer questions. I assume some historic background will be useful from time to time. Feel free to tag me. In case my skillset is needed for larger work items, I am available on a project basis as a freelancer. Just send me a mail.

What is next? Don’t know yet. Probably open source, strongly typed, pushing bytes over the Internet.