compatibility between endpoint and wasm handlers

[amitu]

@AbrarNitk, @sharmashobhit, while we are working on these two different ways for writing custom backend handlers, endpoint proxying and wasm, we want to ensure at a high level that the author of the package that uses a backend implemented in one or other does not have to change anything more than FPM.ftd file to switch from say a WASM based implementation of todo backend with a django based one.

Endpoint vs WASM

Endpoint approach lets people use any language, library, operating system to host their apis. This is really powerful. People can even run services in their VPC etc, so the service has access to private database etc.

WASM on the other hand would run on edge when we have the managed hosting in edge feature. WASM is much "cheaper" on resources, endpoint approach requires a whole VM. Calls from Rust to WASM do not cross process boundary, where as calls from Rust to endpoint is a whole HTTP request, going to a different server.

WASM is fast, but constrained (though the constraints are getting relaxed everyday).

We Expect WASM -> endpoint and vice versa

Once an app is created, some builder may create a better version of the same app, with same interface so it is drop in replacement. Others builders will create because they would want to add better features etc. Some of those features may work within WASM and some would require specific language/OS/library.

We want to ensure our authors can use apps built by any of the builders, and switch between them seamlessly.