Date: Tuesday, 17 November 2026, 14:00 UTC, OpenDev Call #6
Host: Jesse Chejieh (@jessechejieh:matrix.org), Rank II
Agenda: help-center#23
Previous proceedings: 20 October 2026
Recording: Google Drive
Recap: x.com
Alistair Singh / Snowfork opened with the JAM port and bootnode discovery update. The team has its first major PR merged upstream into the SDK, the spec-msg primitives for offchain messaging in polkadot-sdk#12676, which covers version 0.5 of the design. Parity has since released version 0.6, and Snowfork spent the last month updating to it, with the alignment in polkadot-sdk#13759 and further PRs open. The team also now has a usable JAM testnet, which resolves the blocker named at the October call. On bootnode discovery: today every parachain node is also a relay chain node, so it can reach the relay chain DHT, discover another parachain, retrieve its genesis hash and fork ID, connect to that parachain’s own DHT and find its nodes.
JAM has no relay chain DHT, but it does have state: a parachain can write data into JAM storage and another parachain can listen to that storage key. The proposal, designed in polkadot-sdk#12809, is that parachain A writes its bootnodes, fork ID and genesis hash into JAM storage; parachain B listens to A’s key, dials the bootnode, joins A’s DHT and discovers A’s other nodes. As the chat pointed out, the record carries the genesis hash and fork ID, not only the bootnode multiaddresses.
Alistair demoed this on a JAM testnet with collators alpha, beta and gamma and the parachain service, run with a pre-built PolkaJam binary from Parity, launched via Zombienet, with hack PRs on top because JAM is still in development. For now parachain B is pointed at parachain A with a command-line flag; later this will be discovered automatically or fed by governance. In the demo, alpha added itself as a bootnode, gamma picked up the record, dialed alpha, joined the DHT and discovered beta. Asked by Oliver Tale-Yazdi what is stored on-chain, Alistair clarified that the records live in JAM storage, set through the parachain service on behalf of parachains, not in the parachain’s own storage. Asked which JAM node the demo uses, he said it is a pre-built PolkaJam version from Parity, not a community node. Snowfork takes further questions on Element.
Fresh off milestone 1, the Fellowship welcomes new JAM implementer members: Junha Park has been building FastRoll, a Rust client, for about two years, with a pause of a few months, and is now syncing to the latest Gray Paper and targeting the milestone 2 standards. Arjan Zijderveld is working on PyJAMaz, a Python implementation, and is weighing the full node against the light client path for milestone 2. Prematurata Validator has been working on a TypeScript implementation, also for about two years, with the same full node versus light node considerations. Mihail Kirov joined as a JAM implementer as well.
From the floor came the sponsorship question left open in October, and Oliver Tale-Yazdi took it: how rank III members would sponsor lower ranks for retention and promotion. Oliver said there are no concrete ideas yet and the process will have to be figured out informally, by doing it. What sponsorship means is still open, including whether a member failing retention reflects back on their sponsor, in some recursive reputation system, or whether it is simply a list of names; he has so far held off nominating anyone for lack of clarity.
Oliver also reported that his aura authorizer merge request, polkadot-sdk#13453, is still open. It is the first piece upstreamed from the parachain service proof-of-concept repository into the SDK, where the pieces are being cleaned up and made production-ready; it verifies collator signatures, that they are the correct authors of their parachain blocks. The remaining pieces will follow and need changes in the authorship code and in Cumulus.
Oliver then walked through the recent dotUSD incident on Asset Hub. The fault was not in USD itself, which stayed fully backed, with never more USD in existence than backing tether, but in XCM: the withdraw_asset instruction allowed the total issuance accounting to be manipulated, and that accounting feeds the DEX code in the asset conversion pallet, which was loaded with Treasury liquidity. Through bridged XCMs the recorded total issuance of USD was pushed much lower, arbitrage bots on Asset Hub exploited the distortion, the USD price spiked to $26, and the Treasury liquidity was drained. The arbitrage bots made about $1.3 million in total while the attacker account made around $1,000; that they are the same party is not proven, and the attack could have been carried out far more efficiently, in a single transaction rather than through bridges. The latest runtime upgrade readjusted the total issuance, which is now correct again, but the liquidity pools are nearly empty at around $90,000, and a proposal to replenish them may come.
Alexandre Baldé described his follow-up security work on the exploit. His fuzzing in the months before had not caught the root cause, because he had not considered combining asset conversion pallet actions with PSM actions and XCM, plus a bridge; that combination was the missing ingredient, and it will inform his future campaigns. He has a couple of small findings, with a PR coming this week. Oliver added that he runs a Fellowship fuzzing repository continuously on one machine, dispatching arbitrary XCMs and batches of them and checking post-conditions, across 16 threads at roughly 400 operations per second per thread; its coverage has been extended so it should now find this class of issue. Between Oliver’s XCM fuzzer, Alexandre’s grammar-based approach and Brian’s AI-driven search, three different fuzzing methods are running.
Nasihudeen Jimoh presented the kill switch pallet, runtimes#1344, now wired into Asset Hub. Anyone can post a target, such as PSM swaps or XCM asset transfers, by putting up a 10,000 DOT deposit for swaps as currently set, with the deposit parameters open to reviewer input. The deposit runs for a day, and the Fellowship can end it early, refunding an honest post or slashing a dishonest one. The PR is rebased from the earlier kill switch PR, with safe mode removed. Bastian has given an initial review and the comments are being worked through; more reviews are requested, and Oliver said he would take a look.
Next call: Tuesday, 15 December 2026, 14:00 UTC, the third Tuesday as always · agenda thread opens ~1 week out.

