Migration Strategy Details

April 6, 2023

Action Items:

Everyone reviews the risk table and fills in more risks or the mitigations: Risks and Mitigations Table
Everyone reviews Migration Strategy through an IPNS Key
@Gus Eggert: to write a design doc on things that will need to change on the kubo side before we commit to a date
@Gus Eggert and @Guillaume Michel to catch up (on Wednesday) and report back on Thursday on the estimated timeline and effort needed on the implementation front.
@Yiannis Psaras to have a first version of the text for the discussion forum thread/announcement.
Implementation Plan:
- Implementation finished: End of May
- Refactoring of the DHT that could accommodate DH, but without having double hashing: June
- Testing and perf eval through private cluster: June
- Refactoring of the DHT that includes DH, together with the IPNS Key: June
- Wait for enough peers to update before we can turn the IPNS Key: timeline unknown

Risks and Mitigations table
- Draft here: Risks and Mitigations Table
Double Hash DHT (and IPNI) is an EngRes OKR for Q2.
- We need to get rough dates jotted down and commit to them to the extent possible.
- If Q2 is not an option for us with the current capacity, we should flag it now and see what we’d need in order to make it happen.
- @Gus Eggert to sync with @Guillaume Michel and check on the implementation parts
- @Gus Eggert to write a design doc on things that will need to change on the kubo side before we commit to a date.
@Yiannis Psaras: Maybe my concern is invalid, but I feel that If we do the IPNS Key turn based on a date (that we’re going to fix when we see that enough peers have updated), we’re just kicking the can down the road.
- The thing that we gain is that we test the new DHT in production but with a backup and we have the chance to make some comms and give a longer notice.
- The situation might be different if we don’t do it by date, but have nodes upgrade progressively when X% of peers in a node’s routing table have upgraded.
What happens with Hydras? Do they upgrade? Do they turn them off?
- Hydras are on IPNI’s court, so they should be the ones deciding.

We need to figure out the resource usage of three main players. We can start with the resource usage of the current DHT as a benchmark and assume that the double-hashed DHT will perform similarly.
- DRI: @Dennis Trautwein @Anonymous
- DHT Servers → https://protocollabs.grafana.net/goto/jZG1z9BVk?orgId=1
- Content Providers
- DHT Clients
❓ What happens with Bootstrapper nodes?
- DRI: @Guillaume Michel
- Are we going to stop operating the old DHT on bootstrappers at some point? And if so, when?
- They should update first and run both DHTs for some period of time.
From @Gus Eggert: What does the IPNS key structure look like?
- DRI: @Gus Eggert
On March 28, 2023 we discussed having notification text (in the kubo CLI?) to inform users about the number of users remaining in the old DHT. Need to provide more clarification on this.
- DRI: @lidel
  - Digged out the related issue: https://github.com/ipfs/kubo/issues/6487
    - Started work in https://github.com/ipfs/kubo/pull/8839 but never finished (shomatis is no longer with us)
Can we think of any user, main player, content provider, application provider that might not want to switch to a privacy-preserving DHT?
- DRI: @Guillaume Michel