Entity Caching - Router can propagate unnecessary cache-control directives to client

[FineAndDanD]

Describe the bug

I encountered a situation where someone has a custom Java library that intercepts the response from a subgraph before it goes out to the router and inserts a cache-control header into that response.

The problem is:

• With entity caching enabled it seems that it allows subgraphs to insert additional directives programmatically, but the router does not filter out unnecessary values, which is a confusing experience for clients.

The router does disallow invalid values and there does seem to be some logic to validate the directives present (ex. when passing no-cache it changes it in the final response to no_cache)

To Reproduce

1. I've created a reproduction subgraph here, you'll need to clone it and run through the README: https://github.com/FineAndDanD/springboot-dgs-content-cache-repro
2. I've also created a graph in Studio. I didn't want to expose the API key so please feel free to reach out if you want to avoid creating your own graph. Since this is an enterprise feature a graph ref and API key will be needed.
3. Ensure you have a cache running with the basic entity caching router config from here: https://www.apollographql.com/docs/graphos/routing/performance/caching/entity#configure-router-for-entity-caching
4. Start the subgraph, the cache and the router and send a request! I've added a screenshot of what you should see in the headers.

Expected behavior

The expected behaviour is that the router will filter out unnecessary values from the cache-control header when the request resolves. So if there's max-age=10, no-cache, must-revalidate, no-store, the expected value for the cache-control header would simply be no-store. Please see the MDN docs for justification of this reasoning: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Cache-Control#preventing_storing

Output

Desktop (please complete the following information):

• OS: [e.g. iOS] MacOS
• Version [e.g. 22] Sonoma 14.7

Additional context

N/A

[bnjjj]

Thanks @FineAndDanD for all the details and the reproduction steps. I was able to reproduce and I also added a unit test which is currently failing in our codebase. It will be prioritized in order to fix it properly