Include KindVersion in TLE response

[Hayden-IO]

This adds a missing required field for the kind and version of the entry, so that clients will know how to parse the canonicalized body.

This is a much simpler approach to the type factory in Rekor. At this point, given there are only two types, I don't think we should reimplement the factory and simply use case statements.

Ref: #285

Summary

Release Note

Documentation

[codecov]

Codecov Report

Attention: Patch coverage is 80.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 39.33%. Comparing base (7289406) to head (c87e491).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
pkg/server/service.go	50.00%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #286      +/-   ##
==========================================
+ Coverage   39.11%   39.33%   +0.22%     
==========================================
  Files          39       40       +1     
  Lines        2751     2766      +15     
==========================================
+ Hits         1076     1088      +12     
- Misses       1571     1573       +2     
- Partials      104      105       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[cmurphy]

Could you also update switch statements like these

rekor-tiles/pkg/client/write/write.go

Line 121 in 3649260

switch e := entry.(type) {

to use this?

[ramonpetgrave64]

Add tests more tests so we can see the rendered responses

• https://github.com/sigstore/rekor-tiles/blob/364926072f047df0cd4a9c3285c52e2f7f2d5dc9/pkg/server/service_test.go#L189C4-L189C10

[ramonpetgrave64]

I see that KindVersion is also missing from the CanonicalizedBody.

[loosebazooka]

That part is intentional. The type should already be known by then.

[Hayden-IO]

Appu, how is the type derived for monitors reading each entry of the log? With a full bundle, a client will know how to parse it.

[loosebazooka]

I just realized the problem here.

[ramonpetgrave64]

@loosebazooka But isn't the canonicalized by supposed to have all of the values of TLE? That seems to be how it is in v1's responses.

[loosebazooka]

right, I gotta go look at this again.

[Hayden-IO]

Thanks for pointing this out Ramon. We do need entry type information for clients that do not have a Sigstore bundle. I've proposed a fix in #287, which wraps the entry in a one-of so that we retain kind/version information.

Another option Appu and I discussed was trying to replicate the Rekor v1 body with apiversion, kind, and spec. While this might let clients get further along in parsing the body, clients will eventually fail because they won't support the new v2 types.

Appu will be reopening the discussion about a TLE v2 message. Clients will need to update their verification path regardless.

[Hayden-IO]

Also to clarify, KindVersion in the bundle is still needed if CanonicalizedBody, which is optional, is omitted and a client reconstructs the rekor entry from the artifact, signature and verifier. We don't need KindVersion explicitly stated if the message types are retained, which is what 287 proposes.

[Hayden-IO]

Following up on comments in another PR with adding log ID.

[loosebazooka]

oops, stamped this a little too hastily