The validation of therequired keyword does not make sense as it is implemented right now, IMHO.

1. It really does not make any sense to list properties inrequired that do not exist.
2. If you want to use it withallOf and therefore have properties inrequired that do not exist, there are already precautions (feat: handle allof when required linked to parent object (#97) #154).
3. Currently, there is no validation error in the standard case but as soon as I haveallOf in my schema, there is a validation error. That is inconsistent. The PR that breaks it:Allow undefined required properties #312
4. In$referenced schemas, the validation does not take effect either.

One of the main reasons why this should urgently be validated (as was in version 0.7.0) is that validators (that use the specification to validate requests/responses) will fail if there is e.g. a typo in one of the entries inrequired. And now there is no way to detect that.

In general, the OpenAPI/JsonSchema specification is quite relaxed. If only the bare minimum should be validated, this is probably not the only validation that can basically be removed completely.

My take on this:

1. Best case: Throw validation errors for things that are obvious errors (like specifying a field as required that does not even exist).
2. Second best option: make it configurable ("strict mode") or filterable (error IDs to exclude, filter levels, ...).
3. Absolutely necessary minium: make it consistent.

I'd be willing to provide a PR. But since this has been goingback andforth I'd like a decision first, where this is going.