Binary Fuse filters as an alternative to BIP 158 GCS

The recent discussions on the Bitcoin Optech podcast, highlighted by Mike, have brought attention to the concerns about bandwidth usage in light clients more than CPU utilization. This aligns with previous views and feedback from the community, which have traditionally emphasized the performance aspect of these clients. Binary Fuse filters, recognized for their CPU efficiency, continue to be a point of interest due to their speed and design elegance.

In response to the feedback, exploration into alternative solutions to Golomb Coded Sets (GCS) has begun, focusing on Ribbon and BuRR filters as potential replacements. Initial tests conducted over a dataset of 50,000 blocks from the mainnet show promising results for these filters. Preliminary data indicates that both Ribbon-16 and BuRR-16 filters significantly reduce raw filter bandwidth by approximately 20% compared to GCS. However, it's important to note that this reduction in filter size does not directly translate to a proportional decrease in total bandwidth usage for a wallet, as block downloads remain a substantial component of overall bandwidth, especially for larger wallets.

Further examination into false-positive rates for different wallet sizes reveals nuanced performance across the filters. While Binary Fuse 16 shows higher false positives compared to GCS, it remains competitive in terms of speed. On the other hand, Ribbon and BuRR filters, although not as rapid as Binary Fuse filters, still outperform GCS in speed metrics. The comprehensive benchmarks necessary to make definitive conclusions have yet to be performed, but the early data is promising.

Continued testing and refinement of these filters are planned, with more detailed results expected to be published soon. This ongoing research will help determine the practical implications of adopting different filtering solutions for light clients in terms of both efficiency and bandwidth management.