Phase 1 known issues checklist

[hwwhww]

A. Phase 1 beacon chain for legacy

• 1. Possible beacon chain changes for phase 1 Possible beacon chain changes for phase 1 #1825
• 2. Exit fee #1521
• 3. Maximum reversions: Enshrine dynamic weak subjectivity reversion period into fork choice rule ("second finality mechanism"): #1520 and #1053

B. Phase 1 beacon chain for shards

• 1. Shard chain slashing
• 2. Rewards on shards #1280 and #1053
• 3. Implement get_start_shard logic (get_start_shard proposal #1858)
• 4. Slowly rotating shard committees for stabilizing subnets
• 5. Add Shard field to AttestationData
• 6. Use non-length mixins root for shard_data_roots
• 7. Decide how to change active shard count (Shard count recap and discussion #2033)

C. Phase 1 shard chain

• 1. [shard chain fork choice rule] move get_winning_proposal logic to it. Maintain canonical shard chain upon (i) receiving new shard block (ii) new beacon chain block (ii) time tick. (Shard fork choice rule #1773)
• 2. [shard chain transition] Add shard_transition_digest wrapper. Currently we have hash(hash_tree_root(shard_state) + beacon_parent_root + shard_body_root)
• 3. [data availability] Inefficiencies in data availability proofs #1194 (outdated???)
• 4. [data availability] Generating data availability roots from shard block roots #1438
• 5. 0.001 bit proof of custody (0.01 bit proof of custody [depends on 256 bit custody atoms] #1889)

D. Phase 1 fork choice rule

• 1. Separate beacon and shard logic
• 2. Use GHOST instead of LMD-GHOST for shard fork choice rule

[terencechain]

Not sure if this categorize as a beacon chain known issue, I'm curious to know if we should add Shard as a first class citizen in AttestationData, whether to replace CommitteeIndex or be explicit about both

[vbuterin]

Add an item: move to Dankrad's 0.001 bit proof of custody? Also change the custody duration to 8 weeks to avoid the need for complicated delay handling logic; it would be a more simple "if you don't publish the period N key during period N+1 you get forcibly exited".