P2P Networking #4 July 15, 2026

Protocol Calls & happenings
1

Agenda

  • Client updates
  • SPREAD by Matheus

Meeting Time: Wednesday, July 15, 2026 at 15:30 UTC (60 minutes)

GitHub Issue

2

Meeting Summary:

The meeting focused on client updates and a presentation about SPRED, a new gossip sub extension developed by Matheus Franco and colleagues at the University of Lisbon. Daniel from Lighthouse reported significant improvements in network performance, with traffic having halved and plans for a release the following day. Bosul provided updates on Gath implementation progress, mentioning that PRs had been merged to master and work was ongoing on subsequent PRs including tests and protection mechanisms. Matheus Franco then delivered a comprehensive presentation on SPRED, explaining how it addresses the trade-off between sender anonymity and low latency in gossip protocols by implementing a geographic-based approach that uses local random walks combined with efficient long-hop dissemination. The protocol was demonstrated to provide better anonymity and performance compared to both standard GossipSub and Dandelion Plus, with results showing improved latency and reduced attacker accuracy. Participants raised questions about the privacy metrics definition, potential censorship resistance, and concerns about message propagation failures, with Matheus acknowledging these as areas for future research and implementation considerations.

Click to expand detailed summary

Kamil led a meeting that began with client updates, followed by plans for a presentation on privacy improvements in Gossip Sub by Matthias. Daniel reported that Lighthouse has been working on fixes and is preparing to release an update tomorrow, with traffic having significantly decreased since usage increased. Daniel mentioned he might publish a blog post with additional details about the improvements.

The team discussed node observations, with Daniel noting that most observed nodes appear to be super nodes. Bosul provided updates on several PRs, including one for Gath that was merged to master that day, along with ongoing work on subsequent PRs including tests and protection mechanisms. Bosul also mentioned opening a PR for Prism to implement purchase updates and get blobs functionality, which is currently under review.

FLCL discussed switching to test sparse blog posts and mentioned including soft limits in the feature branch. They worked on implementation for disk V5 and V4, which now includes a fork hash to help decline nodes earlier if they belong to different networks. FLCL also mentioned topics as an interesting feature that may speed up discovery in small networks, though Kamil was not familiar with the details and requested more information.

Matheus Franco presented Spread, a gossip sub-protocol extension that aims to improve both sender anonymity and low latency in GossipSub. The protocol uses a geographic clustering approach combined with random walks to balance anonymity and efficiency, offering a spectrum of configurations that can be tuned based on performance or privacy needs. The evaluation showed that Spread improves both latency and anonymity compared to GossipSub, with potential improvements of 20-30% in mean stretch and latency while offering better tail behavior. The protocol is implemented as an opt-in feature in GossipSub PubSub and is available for review and feedback.

Matheus Franco explained two approaches to evaluating anonymity in gossip protocols, including formal mathematical proofs using information theory and differential privacy, as well as practical attack simulations using curious nodes. Kamil clarified that the attacker model involves curious nodes collaborating to identify message sources, with higher fan-out making de-anonymization easier. Matheus Franco acknowledged that while the protocol requires probabilistic random walks for mathematical anonymity, they haven’t yet analyzed potential censorship issues that could arise from long initial low-fan-out paths.

The team discussed concerns about message propagation robustness in the Spread protocol, particularly regarding potential failures when messages don’t reach multiple peers. Matheus Franco explained existing solutions including repropagation mechanisms and noted that current gossip methods already address most problems. Raúl highlighted how attestation propagation could leverage the protocol’s natural feedback loop, suggesting that nodes could test different routes while sending multiple attestations per slot, with the protocol’s ability to handle up to 32-slot delays providing built-in resilience. The discussion also touched on privacy concerns for large message propagation, particularly regarding aggregator de-anonymization in scenarios with large fan-out distributions.

Next Steps:

  • Daniel: Put out a release for Lighthouse, likely tomorrow.
  • Daniel: Consider writing a blog post with more details about the recent performance improvements.
  • Bosul: Work on subsequent PRs for Gossip, including tests and protection mechanisms, to be merged after other clients implement pre-round changes.
  • Bosul: Continue working on the PR for Prism Ripple to implement purchase updated and get blobs before.
  • FLCL: Review and enable sparse blog posts on DevNet 7 next week.
  • matheusfranco: Seek more attention, eyes, and feedback on the SPREAD protocol to motivate its implementation and adoption.
  • Kamil: Measure the impact of malicious nodes not propagating messages on the spread of SPRED.

Recording Access:

3

YouTube recording available: https://youtu.be/EUGRiY9wRD8