Joystream
WIP: Non-mapping writers to query database
So for example, lets say that information such a view counts on videos or channels did not exist in the chain, but in some external system. It would be useful if all of this could be put into a single API such that server side filtering, ordering and pagination based on ths information was possible. The other alternative is for clients to call two separate APIs, somehow reconcile the data, and then do this processing client side, which itself may be quite heavy in terms of processing. Come to thin of it, The Graph must already have somthing like this, as they support exposing IPFS data in their schemas. Now I don't think this writer would need to have access on every block, but not sure what sort of general policy makes sense across possible use cases.
How do we guarantee that the resulting query node is deterministic and all the instances have exactly the same data? With ipfs this is sort of true due to its self-identifying nature, but for other external API I don’t see how to enforce consistency and make it replayable