| Report Date | 2026/07/16 |
| Submitted by | eskimor |
12pRzYaysQz6Tr1e78sRmu9FGB8gu8yTek9x6xwVFFAwXTM8Parachains ConsensusThe designs of the previous period moved into implementation this period - partly by me, increasingly by colleagues building on those designs. My role has shifted accordingly: driving the designs forward, implementing core pieces and reviewing/guiding the implementation work of others.
The low-latency v2 design is now being implemented by a group of engineers, coordinated via tracking issues I authored: low-latency relay chain, low-latency cumulus changes and collator protocol V4.
Core PRs by myself here:
Implementation work building directly on this design (selection, all merged): @iulianbarbu - scheduling info PVF verification and scheduling signature verification, @alindima - older relay parents in prospective-parachains and v3 descriptor support in the runtime, @mchristou - cumulus resubmission store, @serban300 - v3 find_parent() adjustments, @AlexandruCihodaru - v3 descriptor support in the experimental validator. I reviewed most of these.
The speculative messaging design went through three major revisions this period, and is now being implemented: v0.3 - flat per-destination commitments, v0.4 - factoring out off-chain block verification and v0.5 - streams and PoV lifts.
The off-chain block verification design that fell out of v0.4 is a component in its own right, with applications beyond messaging.
Implementation has started, based on my design: @lexnv is working on a PoC, @naijauser on the primitives.
The storage design of the last report got merged and subsequently moved into its own repository: paritytech/web3-storage. It is now an active implementation project with roughly ten contributors - among them @bkontur, @franciscoaguirre, @ilchu and @danielbui12 - covering the storage provider pallet, challenge protocol, provider node and S3 compatible interface. I keep refining the design as implementation questions come up, e.g. challenge clarifications.
Fix coretime partitioning + super low latency on-demand got merged: on-demand orders now appear in the claim queue in the very same block the order is placed, down from ~6 seconds. This also fixed the broken claim queue builder for partitioned coretime.
Following up on the QA learnings from previous periods, I built deterministic simulation based testing for subsystems together with the required clock abstraction. Tests declare the chain as facts and assert only on the observable contract of the subsystem. This makes them:
The framework also supports known-bug tests, enabling a test-first workflow: expected behavior gets merged and pinned before the fix exists. Writing edge-case scenarios became cheap enough that the pilot alone surfaced several production bugs (e.g. claim queue handling at session boundaries). This matters increasingly as AI contributes more code: the contract is pinned by tests that don't change with the implementation.
Resilience and correctness fixes, mostly found while working on the above (all merged, several backported to stable releases): statement distribution pruning bug, rotation bug fix + simplification, enforce current relay parent to be available, report back when candidate is rejected, UMP signal handling cleanup.
Per the Manifesto's requirements for Rank IV (Architect):
"Primary role in ideation and subsequent formalisation of a major protocol component": Low-latency parachains, speculative messaging and Web3 storage - all three designs are formalised in design documents and are in active implementation.
"Primary role in code-design and implementation": V3 candidate descriptor (relay chain and cumulus side) merged and deployed as node feature; deterministic simulation testing framework merged.
| Rank | Activity threshold | Agreement threshold | My record |
|---|---|---|---|
| IV | 60% | 90% | I voted on 60 of 106 referenda I was eligible to vote on in this reporting period (56.6%). I voted against two: #512, which ended up rejected (i.e. in line with the consensus), and #565, which is still being decided. |
Lifetime: 307 votes according to subsquare.