[Quick contemporaneous notes by Ben Edgington; fka "Eth2 Implementers' Call"]
Agenda: https://github.com/ethereum/eth2.0-pm/issues/232
Livestream: https://youtu.be/24MRTRDJ-iw
Different client ratio from the previous devnets: 70% Prysm, similar to mainnet.
A few minor things being worked out. Lighthouse has lost a lot of peers - the team is investigating. Performance has not been as high as expected overall. Lodestar has improved performance by spreading validators across more nodes.
LH is downscoring peers for late sync messages: may lighten up on this.
Do we still expect to upgrade Pyrmont next week? Default is yes, unless any objections. The devnet is finalising, and things look broadly good, so expect to go ahead unless anything turns up meanwhile.
Testing release coming on Monday to cover some edge cases. Prysm found a bug that was not covered by a current case.
For acceptance criteria for Altair on Pyrmont, this doc has some relevant criteria. We should also experiment with turning lots of nodes off, and test non-finalisation, etc.
Dates for Prater and Mainnet? Too early to set these. See how Pyrmont goes first.
Lodestar
Participated in devnet-3. Went well. Working on lowering memory usage; found 3 strategies. Will release a light client article on Monday with a working demo.
Prysm
Mostly focused on merging changes from Altair to main branch. Found a bug in the state transition (which the new test will cover). Running short-term testnets with other clients for Altair - this has been helpful, especially around gossip scoring and fork epoch transition.
Finally impemented all standard API endpoints in develop branch, now working on the v2 endpoints for Altair.
Lighthouse
Published a v1.5.0rc0 pre-release for the Pyrmont upgrade. Stable v1.5.0 will be mainnet targeted after Pyrmont. Includes doppelganger detection. Altair work is merged in. Working on weak subjectivity sync and devops. AWS is very expensive for data transfer for our kind of network.
Grandine
Fixes and optimisations. Full focus now is on multiple runtimes with Altair fork support. A couple of weeks, or maybe a little longer.
Teku
v21.8.0 is out. Has Altair support built in, ready for Pyrmont.
Nimbus
Now have multi-architecture docker images. Released v1.4.2 - update for London, DB migration. Have a dedicated Altair branch. Will provide instructions for those who want to run the Pyrmont upgrade.
Proto
PR for separate block builders and proposers in sharding is up. Creates a market for blocks, and avoids validators participating in every L2. Not yet finalised: please review and comment.
Also reduces networking demands on validators.
The Merge spec has been rebased to Altair and London.
uint256
calculationsHaving the base fee as a uint64
could in principle lead to overflow.
uint256
in the consensus layerDanny likes #2; Mikhail favours #1 (but could live with #2). A uint256
type will be needed for shards in any case, so we might as well do it now.
(Discussion ensues…)
Alternative idea: cap the value of base fee on the execution side.
Note that, for this very limited use case, using uint256
is not computationally expensive. However it will have implementation costs. Eth1 libraries could help.
Specific CLI arguments for handling exceptional scenarios during the Merge. Allows for "social coordination" in the face of miner attacks, etc.
EIP-1011 gives some examples of such CLI settings.
Concept is good, but clients may have different naming conventions. We should aim to keep option names as similar as is reasonable.
See the related issue for further discussion.
There are new BLS test vectors available. Now BLS libraries can be tested outside the Eth2 test suite, and the tests can be used by other projects.
No.