Engine needs to cache JSON schema

[SteveL-MSFT]

Summary of the new feature / enhancement

Currently, for every operation, the engine requests the schema so it can be spawning the same process over and over (or reading it from the manifest over and over). For a set, the engine will perform a test (or synthetic test using get) to validate the input and then the output, then perform the set, validating again, then perform a get to get actual state and validate again. Each validation re-retrieves the JSON Schema

Proposed technical implementation details (optional)

Keep a cache of JSON schema per resource/version. Only retrieve the schema if it's not in the cache. In memory cache only.

[michaeltlombardi]

This is supported in the jsonschema crate with custom retrievers.

As part of the schema canonicalization effort, we want to migrate towards keeping a schema registry for DSC-defined schemas. We could define an in-memory retriever and register various schemas that way. We could leverage the lookup using URI's to construct URIs like:

• resource://Microsoft.Windows/Registry/instance to retrieve the instance-validating JSON schema
• resource://Microsoft.Windows/Registry/filtering to retrieve the export-filtering JSON schema
• extension://Contoso.Example/Secret/options to retrieve the options JSON schema

We would pre-register the DSC schemas and then be able to lookup the externally-provided schemas with the retriever.

[SteveL-MSFT]

Looking at custom retrievers, I'm not sure it actually helps in this case. It looks like it's a convienence feature since the schemas are essentially stored in a hashmap with the $id being the key. In our case, particularly with 3rd party resources, we key off the type name instead since the properties provided in the config generally don't specify the associated $id. The other thing is when DSC engine performs validation, it knows exactly which schema it wants to use already.