The example docker-compose.yml file is available here: docker-compose.prod.yml · main · nodes / Duniter Datapod · GitLab. As you can see, it does not depend on Duniter at all ! (more on that on the topic Datapod : spécifique à un réseau ou agnostique?).
The only components are :
- a kubo node (the image only adds custom peering config that can be seen here: scripts/configure.sh · main · nodes / Duniter Datapod · GitLab)
- postgres + hasura
- the datapod code in typescript
- (pubsub and gateway are not relevant for now)
The only thing that is related to the network is the address format used in the db (the prefix is now 42, common for all tests networks). There are plans to add offchain transaction comments, which could lead to add a dependency to Duniter or Squid in order to allow search in all comments (onchain and offchain), but this will be developed later.
Some advantages of offchain datapods are:
- It allows someone with no write permission on blockchain to write something in the datapods.
- It is much faster than the 6 seconds block time (+finalization) which opens for other features
But we could also design an onchain oplog as discussed in Oplog onchain vs offchain.
The network used is Amino DHT, the public IPFS DHT, it is independent from Duniter p2p network, contrary to the v1 datapods which were based on Duniter 1 p2p network.