It can be observed lately that more and more clients are supporting the use of <tag>@<digest> references for OCI-artifacts in general on their user interface. This naming format conflicts with the OCI-specification, which clearly states that "<reference> MUST be either (a) the digest of the manifest or (b) a tag. The <reference> MUST NOT be in any other format.".
It's clear that the user-client interface is not bound to the OCI-specification. Nevertheless, it would be meaningful to give some guidance on this naming format to ensure consistency across the different levels in the ecosystem.
I see two options:
- Keeping the OCI-specification as it is, but give some guidance on how to deal with it on other levels in the ecosystem.
- Extending the specification to make
<tag>@<digest> also valid, specifying the expected behavior (probably ignoring the part).
@sudo-bmitch apparently regctl supports this format and you are a maintainer of both regctl and this specification. What's your opinion on this topic?
It can be observed lately that more and more clients are supporting the use of
<tag>@<digest>references for OCI-artifacts in general on their user interface. This naming format conflicts with the OCI-specification, which clearly states that "<reference>MUST be either (a) the digest of the manifest or (b) a tag. The<reference>MUST NOT be in any other format.".It's clear that the user-client interface is not bound to the OCI-specification. Nevertheless, it would be meaningful to give some guidance on this naming format to ensure consistency across the different levels in the ecosystem.
I see two options:
<tag>@<digest>also valid, specifying the expected behavior (probably ignoring the part).@sudo-bmitch apparently regctl supports this format and you are a maintainer of both regctl and this specification. What's your opinion on this topic?