Observer and Archive Nodes should be able to support clients with executing script. Fundamentally, the node-client interaction is untrusted, where either party could be actively malicious. This includes
malicious clients trying to attack or spam an honest node with malicious scripts / requests
malicious nodes trying to present wrong results to honest clients
simplified interaction between honest node and honest client with a (limited) trust relationship
In this particular scenario, the client is willing to forgo any proofs and trusts the node. From an engineering perspective, this mode is the easiest to implement with high practical relevance.
In the long-term we want to support two methods of interactions:
Method A (trusted)
simplified interaction between honest node and honest client
However, even in this scenario, an honest node should still be reasonably resilient to malicious clients.
Method B (trustless)