`--alg` should be required when performing signature validation

[mike-engel]

`--alg` should be required when performing signature validation

[mike-engel]

I mostly set a default because it seems like the most common algorithm is HS256, which would allow most users to avoid having to add an extra flag.

Has the industry moved onto a different algorithm by default? Maybe we can error if the algorithm doesn't match, rather than continue to decode? I think that might be a good middle ground solution which helps everyone. What do you think?

[codedust]

From my experience, asymmetric algorithms like PS512 and PS512 are very common (and an absolute must) in scenarios where the authorization server and the resource server are not operated by the same entity. This article gives a great overview over asymmetric algorithms in the context of JWTs.

In general, being aware of the choice of algorithms for signature validation is a good thing. However, setting sane defaults is a good thing, too. With #133, the new warning message provides useful feedback when the algorithms didn't match:

The JWT provided has a different signing algorithm ({:?}) than the one selected for validation ({:?}){/$}

Maybe the actual value of this change is the ability to provide a list of algorithms for signature validation. This is very useful when multiple variations of the same algorithm (e.g. PS256, PS384 and PS512) should be accepted.

To cut a long story short, I think we can also add the HSxxx algorithms as a default choice when #133 is merged. Would that be in your interest?

                      Arg::with_name("algorithms")
                         .help("a comma-separated list of algorithms to be used for signature validation. All algorithms need to be of the same family (HMAC, RSA, EC).")
+                        .default_value("HS512,HS384,HS256")
                         .takes_value(true)
                         .long("algs")
                         .short("A")
                         .possible_values(&SupportedAlgorithms::variants())

[mike-engel]

@codedust that works for me, I think