Jon Johnson

> If #826 allows the server to build a descriptor with the content dereferenced into the data field, then that solves my issue. This is exactly the intention of that...

> It seems to me what you are proposing may end up having an image manifest that looks like this: The example given isn't really a valid image manifest as...

> I had asked about this a while ago when I was first learning about the spec and my understanding was that we couldn't use image manifests for other things...

Going to brain dump some ideas from the OCI call before they're lost to time... As per [this comment](https://github.com/opencontainers/image-spec/issues/803#issuecomment-742088667) we could add a new dimension to manifest lists. Maybe as...

I've run into this myself. This seems like a good idea -- have you brought this up with the runtime-spec as well?

Also torn on this, mildly in favor of it. Especially with the [artifacts stuff](https://github.com/opencontainers/artifacts), I'm not sure if we can rely on the config file to convey this information --...

> Since most images will have fairly shallow reference trees, it should be straightforward for registry code to check for cyclic references. Cyclic references should be [effectively impossible](https://github.com/opencontainers/image-spec/issues/778) since the...

> I am not sure that waiting on IANA is _entirely_ required? Right, I'm not proposing we revert that or anything, but it would be nice to follow the existing...

@dmcgowan definitely, which is why I'm nervous about having multiple suffixes. I'm consdering something like this: ```go func main() { mt := "application/vnd.oci.image.layer.nondistributable.v1.tar+gzip+enc" // OCIMediaType { // Type: "application/vnd.oci.image.layer.nondistributable.v1.tar+gzip", //...

I'm also a bit worried that things like https://github.com/opencontainers/artifacts/issues/11 gaining traction could cement us in a bad spot if we aren't careful. Do these all fit in our world? *...