Consider disallowing additional (not in schema) properties

[daankoning]

I can see some clear merits to either side. The main argument in favour seems to me to be that it would greatly improve ease of parsing if all the attributes can be known ahead of time. On the other hand, different people might have use cases that currently are not covered by the schema, such as different formats or additional data on speeches.

Implementing it would be as trivial as adding some "additionalProperties": false attributes.