Consider the case of importing an RSA-PSS public key using JWK.

WebCrypto callers can request usages of either ['verify' or []  (no usages)

If empty usages are specified then it is possible to import the following JWK:


 {
   "kty": "RSA",
   "key_ops": ["encrypt", "decrypt", "wrapKey", "unwrapKey"],
   ...
 }

The above key is for an encryption algorithm, but it is allowed to be imported for a signing algorithm because the requested usages, [], were a subset of the usages granted to the key.

OK fine.

However, it is NOT possible to import the following:


 {
   "kty": "RSA",
   "use": "enc",
   ...
 }

Conceptually these represent the same kind of key, however WebCrypto enforces "use" differently from "key_ops". "use" in this case is required to be an exact match of "sig".

The same situation applies to ECDH keys.

I believe the key_ops behavior is the correct one, and "use" should work the same way. I feel this way because JWK allows pairing unrelated usages in key_ops (although discourages it) [1]. So it stands to reason that such a key should be importable into WebCrypto.


[1] Section 4.3 of JWK spec says: "Multiple unrelated key operations SHOULD NOT be specified"   .....  SHOULD NOT != MUST NOT