[Authorization] Revert new AuthorizationProvider method

[michaeljmarshall]

Motivation

In PR #12600, we added a method to the AuthorizationProvider that is not used outside of implementations of the AuthorizationProvider itself. Since this interface is public, we should make sure that all methods are necessary. Further, since we cherry picked the commit for #12600 to branch-2.9 and branch-2.8, we introduced a potentially breaking change for third party implementations of the AuthorizationProvider interface. While the commit itself isn't breaking, it does imply that the interface has a method that could be used by the authorization service or some other part of Pulsar. Since third party implementations might not have this method implemented, they could break if we started to use the method. Since #12600 includes an important bug fix, we should not couple an expansion of the API with a bug fix.

For added context, it is likely that this method was added because there are other similar methods in the interface. For example, the interface has the following methods: allowSinkOpsAsync, allowSourceOpsAsync, and allowFunctionOpsAsync. These methods are much older and are called by the AuthenticationService.

Note: this was only released in 2.9.0. It hasn't been in any other releases yet.
Note: the new method has not yet been published in any releases. It is set to go in 2.8.2 and 2.9.1.

Modifications

• Remove the AuthorizationProvider#allowConsumeOpsAsync method from the interface and from all implementations of the interface.

Verifying this change

This is a trivial change. No underlying logic is changed by this commit.

Does this pull request potentially affect one of the following parts:

This fix addresses a potentially breaking change.

Documentation

• no-need-doc

[lhotari]

Good catch. LGTM

[michaeljmarshall]

/pulsarbot run-failure-checks

[michaeljmarshall]

/pulsarbot run-failure-checks

[michaeljmarshall]

/pulsarbot run-failure-checks

[codelipenghui]

Note: the new method has not yet been published in any releases

I think the 2.9.0 already contains #12600? do you need to also revert the 2.9.0 release?

[michaeljmarshall]

Note: the new method has not yet been published in any releases

I think the 2.9.0 already contains #12600? do you need to also revert the 2.9.0 release?

@codelipenghui - you're right, it was included in the 2.9.0 release. I thought it had missed the release because it has the label release/2.9.1. I'm not sure what the right choice is for branch-2.9. The release is not good to use in production, so perhaps it is okay to remove it? If it is not okay to remove it, we could annotate the method as deprecated. What do you think we should do?

[freeznet]

LGTM

[codelipenghui]

@michaeljmarshall I think it's ok to remove it in 2.9.1 since we will not introduce breaking changes to 2.9.1, just a user implemented method will not be used. So we only have a breaking change in 2.9.0. /cc @merlimat Please help check.

[eolivelli]

Lgtm