The SHA3 OIDs are deprecated

[Jakuje]

The x509 SHA3 OIDs listed here:

cryptography/src/rust/cryptography-x509/src/oid.rs

Lines 130 to 137 in 5e73a91

	pub const SHA3_224_OID: asn1::ObjectIdentifier =
	asn1::oid!(1, 3, 6, 1, 4, 1, 37476, 3, 2, 1, 99, 7, 224);
	pub const SHA3_256_OID: asn1::ObjectIdentifier =
	asn1::oid!(1, 3, 6, 1, 4, 1, 37476, 3, 2, 1, 99, 7, 256);
	pub const SHA3_384_OID: asn1::ObjectIdentifier =
	asn1::oid!(1, 3, 6, 1, 4, 1, 37476, 3, 2, 1, 99, 7, 384);
	pub const SHA3_512_OID: asn1::ObjectIdentifier =
	asn1::oid!(1, 3, 6, 1, 4, 1, 37476, 3, 2, 1, 99, 7, 512);

are deprecated values that should not be used anymore:

https://hosted.oidplus.com/viathinksoft/?goto=oid%3A1.3.6.1.4.1.37476.3.2.1.99.7.384

The NIST values {joint-iso-itu-t(2) country(16) us(840) organization(1) gov(101) csor(3) nistAlgorithms(4) hashalgs(2) sha3-384(9)} should be used instead, wherever possible.

If these are strictly necessary for some use case you need, they should have at least warning about their deprecation and there should be generally usable variants with standard OIDs.

[Jakuje]

This diff worked for me:

 pub const SHA3_224_OID: asn1::ObjectIdentifier =
-    asn1::oid!(1, 3, 6, 1, 4, 1, 37476, 3, 2, 1, 99, 7, 224);
+    asn1::oid!(2, 16, 840, 1, 101, 3, 4, 2, 7);
 pub const SHA3_256_OID: asn1::ObjectIdentifier =
-    asn1::oid!(1, 3, 6, 1, 4, 1, 37476, 3, 2, 1, 99, 7, 256);
+    asn1::oid!(2, 16, 840, 1, 101, 3, 4, 2, 8);
 pub const SHA3_384_OID: asn1::ObjectIdentifier =
-    asn1::oid!(1, 3, 6, 1, 4, 1, 37476, 3, 2, 1, 99, 7, 384);
+    asn1::oid!(2, 16, 840, 1, 101, 3, 4, 2, 9);
 pub const SHA3_512_OID: asn1::ObjectIdentifier =
-    asn1::oid!(1, 3, 6, 1, 4, 1, 37476, 3, 2, 1, 99, 7, 512);
+    asn1::oid!(2, 16, 840, 1, 101, 3, 4, 2, 10);

I can submit a PR if this sounds like a good direction.

[alex]

Hmm, we've got three problems here:

1. on parse we need to recognize both versions
2. on generate, we should use the new OIDs
3. we also have public Python APIs for these OIDs (HashAlgorithmOID.SHA3_224 etc.), which are documented as having these particular OIDs. In these cases I'm not sure what the right behavior is.

@reaperhulk thoughts?

[reaperhulk]

I agree on one and two. For number three I see two options:

1. Add the new OIDs with new names and leave the old forever. Assume that we're fine with SHA3_224_NIST or similar as the "preferred" OID.
2. Add the new OIDs with new names and do a deprecation dance to rename the old SHA3_224 and ultimately move the new OID to that namespace.

I'm kind of indifferent here so maybe option 1 is actually fine. Certainly easier, and we can document which OIDs are preferred.