Deployment notes for v 2.0.5

[HenrikJannsen]

With the changes in UserProfile, AuthorizedProofOfBurnData and AuthorizedBondedReputationData we have to deal with some requirements to limit backward compatibility issues.

For the UserProfile case I tested all scenarios with seenode and app at old version and new versions. The only combination where we have problems is if there runs only a new seed node and a new user profile gets created while the old version app is offline. When going online the old version app gets a failure at the InventoryResponse because it contains the version 1 objects for the UserProfile and therefore it will not get the newly added userprofile.
If staying online republished userprofiles (every 3 hour) get broadcast and the broadcast works as we send both versions (new version fails but old is recognized).
We can mitigate that problem easily by keeping one seed node running on the old version, thus serving the old nodes.
As seed node selection is random there is no guarantee though that nodes will get all user profiles. As it only affects new user profiles the worst is that users on old versions might miss messages from new users.

We could potentially filter the Inventory response to not include version 1 objects, but the limited disruption does not justify the extra effort, complexity and risk.

With #2349 from 2.0.5 on such issues would not happen anymore.

With the AuthorizedProofOfBurnData and AuthorizedBondedReputationData the situation is basically the same. Also here we should keep running an old seed node to serve the old nodes. We also keep one oracle node on the old version to be on the safe side, though as we publish both versions that should not really be needed.

Further planned improvements are to add the Bisq version to the Capability. This would allow us to tailor responses in case we run in future into more issues.

[HenrikJannsen]

See also https://github.com/bisq-network/bisq2/blob/main/docs/dev/backward-compatibility.md

Recent changes removed the requirement to run an old seed node at launch. But it still require more in-depth testing and it might be still good for unexpected issues to leave one seed (and oracle) on the old version until most of the users have updated.