`Message.id` is unclear

[tmillr]

Suggestions

Improve the description for Message.id. This property is required when responding to a request with an error response and the body is provided, but the current description refers to it both as an identifier for the message and an error ID. I assume that it wants an error id/number (e.g. that would be reused later if the same type of error occurs again)? Because a message id is already provided via seq. But even then, is it an id for the instance/appearance of the error, or, for the type of error (like errno in C)?

Also, the description for seq could probably be improved as well. E.g. it could explain that, for responses, it is not required for it to match the seq of a corresponding request (this is somewhat unintuitive and is different from json rpc for example).

It would also help newcomers to specify in the docs that the protocol is distinct from json rpc (even though it may look similar).

[connor4312]

Message.id is entirely adapter-defined. It's up to you want to give. You can have unique numbers for errors, you can have all errors be id 0, you could have a random number created each time. The purpose is to help debug adapters authors track bugs back to their source. Suggestions welcome on the wording.