Namespace the digest tags #68
Conversation
If a similar mechanism is required for other types of references in the future this will allow for disambiguation. Signed-off-by: James Hewitt <james.hewitt@uk.ibm.com>
Jamstah
requested review from
dmcgowan,
imjasonh,
jdolitsky,
jonjohnsonjr,
justincormack,
lachie83,
michaelb990,
nishakm,
SteveLasker and
sudo-bmitch
as code owners
| @@ -171,10 +171,10 @@ Clients SHALL NOT expect manifests uploaded before the [referrers API](#referrer | |||
| For registries that do not support the `referrers` API, a tag MUST be pushed for any manifest containing a `refers` descriptor with the following syntax: | |||
|
|
|||
| ```text | |||
| <alg>-<ref>.<hash>.<type> | |||
| ref.<alg>-<ref>.<hash>.<type> | |||
There was a problem hiding this comment.
Do you have an example of another type of prefix you imagine? I'm a little worried about prematurely generalizing here, if we don't know what else will end up using this.
There was a problem hiding this comment.
I was thinking really about there already being two mechanisms for referrers (oras and the reference types wg), and that this could be a fallback mechanism for either. How would I tell which tag was from where? That is hypothetical.
However, it also helps people who don't know that tags are being used for references have some idea what they're looking at, especially if we take off the type as part of #65.
There was a problem hiding this comment.
I really genuinely hope that when this working group delivers its changes/recommendations to the OCI specs, that there is no need for ORAS to go off and build their own thing. Many ORAS folks are actively involved in OCI and in this WG, so if they have other needs that aren't met by this, they have a seat at the table to change that.
As for other non-ORAS groups that may want their own digest-tag-style disgusting hack spec enhancement, we can evaluate their proposed solution based on their needs and on the state of registries in the wild at that time. Worst case, they add :cats.sha256-abcd as their special namespace, and this WG's type is the "default". But even that I hope never happens.
There was a problem hiding this comment.
Fully agree that I hope it never happens.
What about the indicator to people wondering "why are there all these tags in my repository that I don't recognise?"
There was a problem hiding this comment.
I'm not sure that a ref. prefix does much to answer that question on its own. A person surprised to find :sha256-abcd in their tag list would probably google it and (hopefully, eventually) find a useful answer, just the same as they'd google [ref.sha256 docker tag].
Hopefully even-more-eventually I'd hope registries would adopt support for references and we can be rid of digest tags entirely, some time around 2060. 👴
There was a problem hiding this comment.
Well yes, this is really attempting to put lipstick on a pig, as the saying goes. Will close this one then.
|
No appetite for this. |
If a similar mechanism is required for other types of references in the
future this will allow for disambiguation.
Signed-off-by: James Hewitt james.hewitt@uk.ibm.com