Historical curiosity: Why did we delete half the spec, again?

[jonjohnsonjr]

I cannot for the life of me remember why this happened: c90b0f1

There is a ton of stuff missing from distribution-spec, still.

[vbatts]

It did improve readability. The spec carried over from Docker had a lot of generated cruft.
But it was a huge endeavor. I wouldn't doubt that things were missed.
What particularly are you finding missing?

[corhere]

I, for one, am missing the historical context and decision making behind many of the nontrivial spec changes, such as

- Actionable failure conditions, covered in detail in their relevant sections, are reported as part of 4xx responses, in a json response body.
- One or more errors will be returned in the following format:
+ A `4XX` response code from the registry MAY return a body in any format. If the response body is in JSON format, it MUST
+ have the following format:

and how Content-Type: application/octet-stream became a required request header when PATCHing blobs. Before #178 the spec included a table of headers and parameters that SHOULD be included, which did not include the Content-Type header. The Content-Type header was only mentioned in an illustrative example.

Also:

At one point we deprecated catalog, but I can't seem to find reference to it anywhere after #178

[jonjohnsonjr]

What particularly are you finding missing?

I can't find any reference to the pagination mechanism for tag listing. I think this got dropped when catalog was [re]moved?

It makes a lot more sense to me now that folks had started inventing new pagination mechanisms for the referrers API: they didn't have the existing mechanism to copy!

[sudo-bmitch]

We also define a chunked upload, but not the streaming upload for blobs.

[vbatts]

On 21/07/23 07:41 -0700, Jon Johnson wrote: I can't find any reference to the pagination mechanism for tag listing. I think this got dropped when catalog was [re]moved?

open a PR to add it back! pagination was useful, and if folks have already implemented it, then we ought not introduce new and different ways of doing it.