Replies: 4 comments 7 replies
-
This would also be great for resolving buildpacks/pack#814, while we decide what public interface we'd like in imgutil |
Beta Was this translation helpful? Give feedback.
-
A brain dump of something I've been thinking about recently. Credit goes to the GGCR strategy. 🎤💧 Note that I omitted your projecttoml-parser @jabrown85 based on the conversation in WG today that it's more of a lifecycle concern.
|
Beta Was this translation helpful? Give feedback.
-
If we do this, that would be a great time to systematically take care of buildpacks/pack#664 |
Beta Was this translation helpful? Give feedback.
-
Relevant issue: buildpacks/pack#761 |
Beta Was this translation helpful? Give feedback.
-
As we discussed in the sub-team sync recently, it might be useful to have
pack
broken up into smaller binaries/packages. Like ggcr.Starting off with some ideas based off of actual use cases I've wanted recently.
pack
remains the sameprojectoml-parser
new binary/package to readproject.toml
and produce inputs forlifecycle
(order.toml
,/platform/env
files, etc.)buildpack-fetcher
given a list of buildpack references (or maybe anorder.toml
?) - it would resolve the buildpack references, download to a specified location if they do not exist on the builder, and validate the buildpacks are valid for the builder. The reason this would be useful in a project likepack
is becausepack
end-users will expect non-pack platforms to resolve downloading of buildpacks in the same way as it does for them locally. This is especially true when buildpacks are provided in the source code viaproject.toml
. This being in it's own image would also make tekton more powerful as a pre-step to download additional buildpacks not provided in the builder image.Without at least having some of this exposed as go packages, platform authors will need to try to keep up with any
pack
changes and the platforms may diverge in a way that is undesirable for the buildpack ecosystem.Beta Was this translation helpful? Give feedback.
All reactions