| title | discussion |
|---|---|
Digest input clarification |
{: .no_toc }
{: .no_toc .text-delta }
- TOC {:toc}
The input to digest computations is not clear. For example:
-
The ImageID is the hash of the config JSON, but JSON is text and hash algorithms take bytes, and text has multiple serializations to bytes.
-
For composite objects like
.tar.gzfiles, sometimes the hash is of the compressed object but sometimes it's the uncompressed object. -
Why both "Layer DiffID" and "layer digest"? Don't these serve the same purpose? And if they do have different purposes, the spec should explain the reasoning.
(The Charliecloud team had a lot of trouble debugging this because when a digest computation is wrong, you just get the wrong output with no indication of where the mistake is.)
Whenever a digest is introduced by a spec, it should state in the same paragraph as the introduction (or a similar close location) what the input is.
This suggestion would be relevant to any interactions with digests.