Closed displague closed 1 year ago
/cc @invidian (for Terraform provider relevance)
Is metadata
intentionally a string to stay unbiased about the metadata format?
Shouldn't Tinkerbell have a bias on the metadata format (a bias to follow hegel)?
This would be helpful for adding cloud-init
support for the Tinkerbell metadata service.
@mmlb is it part of the proposal you raised around instance? Can you help us to figure how if @displague has a reasonable request or if it is like it is for good reasons?
packetMetadata.String()
does not exist anymore. @displague if you feel this is still needed, please reopen.
Expected Behaviour
Metadata (from the metadata package) is typed within the Hardware type used by the Hardware client.
Current Behaviour
Metadata (from the packet package) is a string within the Hardware type used by the Hardware client.
As a Tinkerbell client implementor, creating a Hardware object, it is not expected that the client interface should require a string interpretation of the
Metadata
type from theprotos/packet
package.Possible Solution
This may create a bc-break, so the PR should include this label.
Context
I'm investigating a Crossplane client for Tinkerbell (https://github.com/displague/crossplane-provider-tinkerbell/commit/cc989d6dbbe5e5ca16711f44b1a5c9b80a878ddf#diff-261eb121ddd5a8cf59bab09ef855945be48c5730d2f4687a1ee8575e8675eaa0R171-R192).
This would also benefit the Tinkerbell Terraform provider, which currently passes json and yaml blobs around (https://github.com/tinkerbell/terraform-provider-tinkerbell/blob/28bd5d0/tinkerbell/resource_hardware_test.go#L25-L93).
Your Environment
Go importing tink at 384fe30ce2a4