Skip to content

MILESTONES.md

Latest commit

 

History

History
36 lines (21 loc) · 2.59 KB

MILESTONES.md

File metadata and controls

36 lines (21 loc) · 2.59 KB
Raw

Porch — Milestones

Build order. Each version is independently useful and adds exactly one capability. Payments are last (V3).

Phase 0 — due diligence (no code)

  • RPCh post-mortem — the closest prior product (paid private RPC over an incentivized relay) stalled; understand why before committing to the V3 payment layer.
  • Portal Network post-mortem — closest decentralized-state substrate; no incentive layer.
  • Demand check — because payments are V3, the demand test happens inside V0/V1: will wallets route to Porch nodes for free, on decentralization + censorship-resistance alone? If not, payments never matter.

V0 — decentralized serving (free)

Central-server registry + node serves the allowlist on a confined port; a wallet points at a node's URL (or the server's list). No payments, no privacy, trust-based correctness.

Done when: a wallet pointed at a Porch node gets a correct eth_getBalance, and the node self-registers and appears in the directory. (A runnable toy of exactly this already lives in ../demo.)

V1 — on-chain discovery + correctness

  • On-chain registry contract: nodes register (operators pay their own gas); the wallet reads the contract directly to discover. Optional non-slashable stake-to-be-featured.
  • Correctness: the node returns the answer + a Merkle proof (eth_getProof); the wallet verifies it with its own light client (Helios / lightspan). Privacy-neutral.

Done when: a wallet discovers a node via the contract and rejects a tampered answer because the Merkle proof fails against its light-client header.

V2 — privacy (Origin + Content)

  • Origin: anonymizing transport (Tor / mixnet / OHTTP) + byte-identical request canonicalization + reproducible client.
  • Content: PIR (companion spec) — the node can't see which slot/address you read.
  • Research: verifiable PIR — proofs that don't reveal the queried index (PIR + correctness together). The key open problem here.

Done when: two users' identical requests are byte-identical on the wire, and a PIR read returns correct data without the node learning the index.

V3 — payments (last)

Shared vault (fixed-denomination deposit, unlinkable redemption); batch-voucher (default) + nanopayment rails; spent-flag; low governable fee. Slashing turns on — client full-RLN (separate bonded) + node signed-head/quorum, cryptographic proof only, never complaints.

Done when: one deposit funds ≥1k calls; only correct settlements land on-chain; a double-redeem bounces; payments are unlinkable to the deposit.