OCI Annotation Propagation (Base64 vs Plaintext)

[Nachiket-Roy]

bunny uses Base64 for OCI annotations but not always for Docker config labels, resulting in mixed cases in urunc.json or OCI specs. Currently, urunc forcefully calls base64.StdEncoding.DecodeString() on these values, causing container creation to crash if the values are actually plaintext.

[cmainas]

Hello @Nachiket-Roy ,

this is intentional. Latest changes in urunc have allowed us to choose plaintext, however this needs to be done in a controlled way to keep backwards compatibility. The work for #12 is related.

[Nachiket-Roy]

@cmainas Thanks for the clarification, I'll close the PR.
Could you clarify what the expected controlled transition should be? I have a corresponding bunny PR(53) ready, but I'd like to understand the preferred approach.

[cmainas]

We have several Sample images which we need to bring up to date. Furthermore, there has to be an announcement about this breaking change and a period for people to adjust (usually one release).

[Nachiket-Roy]

I assume rebuilding and pushing those will be done by the maintainers? should i revive the PR with legacy fallback + deprecation notice, also bunny pr is already there.