Hey @dholth thanks for pushing this further. Since we're reviewing the awesome implementation in rattler, I was wondering: we're storing two blake2 hashes now, and don't need to exactly match up the repodata.json formatting because of that?

Yes. Although we could format repodata.json in exactly the same way as conda-index, that formatting is not guaranteed. Instead, we store the hash that the server found when creating the patches, and the hash of what we wrote to disk. We save time by not reformatting and the server can change its formatting.

For example the anaconda.org channels don't pretty-format (but also don't have a way to provide jlap diffs).

We could consider adopting a formal json canonicalization scheme in the future.

Thanks for the clarification. I think it'd be nice if the hashes match up but maybe we can fix that another time? Idk

Yes, we could do that in the future but in the meantime we are allowed to change this code https://github.com/conda/conda-index/blob/main/conda_index/index/__init__.py#L825-L833 without breaking clients.