Originally created by @eli-xciv on GitHub (Mar 21, 2024).
It would be nice to be able to store packages in and use packages from an OCI registry.
This issue/request is to support the oci:// URI in addition to package:// for dependencies.
As for OCI artifacts and layer media types, pkl could adopt vendor based media types for each layer, if each generated artifact was added as a separate layer, such as application/vnd.apple.pkl.metadata.v1.xml for the metadata artifact application/vnd.apple.pkl.metadata.v1.checksum_sha256 for the metadata checksum application/vnd.apple.pkl.package.v1.zip for the actual package artifact
and application/vnd.apple.pkl.package.v1.checksum_sha256 for the package checksum
This would possibly also require a change to the pkl-cli to add a flag to dictate how the package will be build (default | OCI)
Originally created by @eli-xciv on GitHub (Mar 21, 2024).
It would be nice to be able to store packages in and use packages from an OCI registry.
This issue/request is to support the `oci://` URI in addition to `package://` for dependencies.
As for OCI artifacts and layer media types, pkl could adopt vendor based media types for each layer, if each generated artifact was added as a separate layer, such as
`application/vnd.apple.pkl.metadata.v1.xml` for the metadata artifact
`application/vnd.apple.pkl.metadata.v1.checksum_sha256` for the metadata checksum
`application/vnd.apple.pkl.package.v1.zip` for the actual package artifact
and
`application/vnd.apple.pkl.package.v1.checksum_sha256` for the package checksum
This would possibly also require a change to the `pkl-cli` to add a flag to dictate how the package will be build (default | OCI)
We've thought about this! It's also how CUE and KCL decided to distribute their packages.
This is somewhat compelling; something that package: is missing right now is some sort of index. OCI registries solve that problem for us.
But, I think we need some more time to understand the trade-offs of our current package model before adding any changes here.
@bioball commented on GitHub (Mar 21, 2024):
We've thought about this! It's also how CUE and KCL decided to distribute their packages.
This is somewhat compelling; something that `package:` is missing right now is some sort of index. OCI registries solve that problem for us.
But, I think we need some more time to understand the trade-offs of our current package model before adding any changes here.
Since I'm looking at using Pkl as a replacement for Helm, being able to reuse our Helm registry for Pkl packages is something of a requirement. Currently considering using oras and some jank around local packages once they're downloaded to work around this for now.
@bvalyou commented on GitHub (Apr 9, 2024):
Since I'm looking at using Pkl as a replacement for Helm, being able to reuse our Helm registry for Pkl packages is something of a requirement. Currently considering using [oras](https://oras.land/docs/quickstart) and some jank around local packages once they're downloaded to work around this for now.
@bioball Just checking back in to see if you all have been able to have any conversations around packaging and possible support for this.
Thanks for any info!
@eli-xciv commented on GitHub (Nov 5, 2024):
@bioball Just checking back in to see if you all have been able to have any conversations around packaging and possible support for this.
Thanks for any info!
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @eli-xciv on GitHub (Mar 21, 2024).
It would be nice to be able to store packages in and use packages from an OCI registry.
This issue/request is to support the
oci://URI in addition topackage://for dependencies.As for OCI artifacts and layer media types, pkl could adopt vendor based media types for each layer, if each generated artifact was added as a separate layer, such as
application/vnd.apple.pkl.metadata.v1.xmlfor the metadata artifactapplication/vnd.apple.pkl.metadata.v1.checksum_sha256for the metadata checksumapplication/vnd.apple.pkl.package.v1.zipfor the actual package artifactand
application/vnd.apple.pkl.package.v1.checksum_sha256for the package checksumThis would possibly also require a change to the
pkl-clito add a flag to dictate how the package will be build (default | OCI)@bioball commented on GitHub (Mar 21, 2024):
We've thought about this! It's also how CUE and KCL decided to distribute their packages.
This is somewhat compelling; something that
package:is missing right now is some sort of index. OCI registries solve that problem for us.But, I think we need some more time to understand the trade-offs of our current package model before adding any changes here.
@bvalyou commented on GitHub (Apr 9, 2024):
Since I'm looking at using Pkl as a replacement for Helm, being able to reuse our Helm registry for Pkl packages is something of a requirement. Currently considering using oras and some jank around local packages once they're downloaded to work around this for now.
@eli-xciv commented on GitHub (Nov 5, 2024):
@bioball Just checking back in to see if you all have been able to have any conversations around packaging and possible support for this.
Thanks for any info!