Last version updated draft
Summary changes:
- return back to Golomb coding
- implemented more simple and effective shema
- Total filters size is smaller then BIP 158 at all total estimated savings more than 20% (exactly info will be soon)
- filter is deterministic and could be committed as commitment in coinbase transaction in future
- flexible GCS parameters to to maintain the necessary FPS
- spliting filter for 2 parts: unique elements and duplicated elements
- duplicated elements could be encoded more effective
Open questions:
- Optimal range for batch?
- Why we need sip has instead of just use first 64 bits from pub key/script hash?
- Downloading unique/duplicated elements separately? Just add filter types for these purposes?
Thanks for any feedback or discussions
Aleksey Karpov