On 3/12/18 4:39 AM, Weijun Wang wrote:
I put "SHA-1" in a DisabledAlgorithmConstraints, it rejects SHA1 but allows sha1.
That sounds like a bug.
The reason is that http://hg.openjdk.java.net/jdk/jdk/file/6b54e8cd9b3d/jdk/src/java.base/share/classes/sun/security/util/AlgorithmDecomposer.java#l96 does not see "sha1". On the other hand, it rejects both "SHA-1" and "sha-1", because it's a direct case-insenstive match. Also, it allows both "SHA" and "sha" because there is no special code for it. Isn't "SHA" also an alias of "SHA-1"? Do you think all these names should be recognized? Shall we clarify it in the spec?
I would tend to think that we should only specify (or guarantee) that standard names are checked and used in the disabled algorithm properties. Aliases have never been a supported/standard feature, so I think if we start to accommodate them, then we have to document that and it increases the complexity of the code and chance that we might miss one. For example, SHA is an alias for SHA-1 in the JDK Sun provider but there is no corresponding alias for SHAwithRSA in the SunRsaSign provider, so it is inconsistent already.