Just to double check:
From a dependency inversion side, "the shader cache could use some interface" is something I interpret as meaning:
  - //gpu code defines some interface for shader caching
  - somewhere 'above' //gpu there exists a //net-dependent implementation (such as //content or //chrome) that implements that //gpu interface
  - somewhere 'above' //gpu (such as //content or //chrome), the //net-dependent implementation is injected into the //gpu code (such as by passing that interface to //gpu initialization code)

That doesn't address the //net->//gpu dependency that exists (indirectly), due to //net knowing that //gpu (or the future //content interface) is a consumer of the disk cache for disk-cache related stats, but that's more an issue for //net needing a 'stats' interface for the disk cache (so that such a //content or //chrome interface could know about different disk cache consumers)

Is that reasonably accurate?

What's not clear to me is where the "Something above //gpu and above //net but knows about both //gpu and //net" is. My understanding is the code originally lived in //chrome, then //content (both of which know about this), but was moved into //gpu due to non-//content consumers. Putting it in //components seems wrong (given the goal of //components), but is instinctively appealing.

That is a correct explanation of what I tried to describe, yes.

One of the motivations for avoiding the direct dependence/use of disk_cache is that using a mojom service like indexeddb/filesystem would possibly allow the gpu-host to be sandboxed more than it currently is.

The //net dependency currently is in //gpu/ipc/host, which is a standalone target (in the sense that no other code in gpu depends on this code). So it could have lived elsewhere too (e.g. in //components/gpu_host). But I couldn't come up with a better location.

This issue has been Available for over a year. If it's no longer important or seems unlikely to be fixed, please consider closing it out. If it is important, please re-triage the issue.

Sorry for the inconvenience if the bug really should have been left as Available. If you change it back, also remove the "Hotlist-Recharge-Cold" label.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot