You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We should add an option for subgraph queries to await at least 2 matching response hashes (for the same request hash) from separate indexers. The primary use-case for this is for internal queries, such as network subgraph queries made by the gateway. For the internal case, we should also add an allowlist of 1 indexer per core dev for increased trust and bootstrapping the network state on startup.
Secondarily, this is an interesting feature to surface to users. It would be informative to see if some users are willing to pay extra for inline cross-checking on queries.
I think this will require a substantial amount of work to support, based on some initial thoughts:
The client query handler needs to optionally await multiple responses, and optionally rerun indexer selection without candidates that have already contributed successful responses to the set. Some design work is needed to avoid a big pile of spaghetti.
In general, we need to fully transition to queries to subgraphs on the network and remove our reliance on the hosted service. This is important regardless because it is a single point of failure for the effectiveness of the gateway that has caused issues in the past.
The text was updated successfully, but these errors were encountered:
We should add an option for subgraph queries to await at least 2 matching response hashes (for the same request hash) from separate indexers. The primary use-case for this is for internal queries, such as network subgraph queries made by the gateway. For the internal case, we should also add an allowlist of 1 indexer per core dev for increased trust and bootstrapping the network state on startup.
Secondarily, this is an interesting feature to surface to users. It would be informative to see if some users are willing to pay extra for inline cross-checking on queries.
I think this will require a substantial amount of work to support, based on some initial thoughts:
The text was updated successfully, but these errors were encountered: