Future of SPV tech
issues with current SPV wallets
Existing clients and their status
- bitcoinj
- electrum they are doing SPV proofs bloom filters
utxo commitments need to get in need to pick which type of utxo commitments
Number of nodes to connect to for more SPV clients
1 header from each peer to verify that we are all getting the same thing what block does each peer think is the best? very cheap to poll a huge number of peers
“commit to a bad block”
- is SPV completely broken? privacy failures with bloom filters
- commit to a bloom filter, only ask for the transactions you want
- you can use pynchon gate and riposte for messaging (like anonymous remailers)
what’s the benefit of the SPV model over the electrum model?
why would a node want to communicate with SPV nodes at all?
off-chain contracts for SPV, pay per query
storageless full-nodes (full verification, but no blockchain storage)
make each node have random fraction of utxo set
bitcoinxt getutxos does not have proof, sybil insecure
spents vs unspent - use unspent, otherwise super large
- order by fee per byte is for block header somewhere, not utxo set commitments
- for commitments to occur, need order
- soft-fork, but not for headers
- utxo commitments don’t have to be in the blockheader miners must only optionally include it if they don’t include it, it’s fine
how to order utxo data before merkleizing it
take historical utxo set, then make updates in same way as everyone else rules for when to merge the tree updates
headers should have a commitment to other headers in the coinbase transaction output
- utxo commitment - tree, leafs should be sorted by outpoints reduce resource requirements for tree updates also maintain a set of tree update deltas ** large pile of tree update deltas should be merged into the actual tree
SPV fee estimation worried about high fees, not low fees (assuming replace-by-fee)
SPV handling of lightning network and payment channels might require utxo set commitments
what’s the purpose or use case of SPV
who funds spv wallet development? does anyone? who might?
lower infrastructure requirements, don’t have to run giant third pirty central server
riposte - private information retrieval, reversal of pynchon gate model
Bitcoin Core should have an SPV implementation there’s a branch for SPV start a full-node, throttles verification why should this be a separate protocol, should be default part of protocol
Summary points
- utxo commitments need to happen, soft-fork
- headers should have a commitment to other headers in the coinbase transaction output
- updatable utxo commitment trees for spv and other reasons
- there should be a SPV implementation in core
- talk to more SPV clients, 1000s of connections
- fee estimation - should you do anything at all?
- lightning network + spv for your phone or other low-powered devices
- argue for spv development for low infrastructure requirements
- pay per query for SPV
- bloom filters don’t work in SPV, denial-of-service problems, need private information retrieval protocol like riposte