Hi Chromium Dev Team,
Recently, I'm researching into Chromium’s architecture for AI Assistant Feature, I have been studying both GLIC (`chrome-untrusted://glic/` with `glic.mojom` / `glic_api.ts`) and the experimental Extension Actor API (`chrome.experimentalActor` in `chrome/common/extensions/api/experimental_actor.idl`).
I would greatly appreciate your insights on two architectural questions:
1. Roadmap & Intent of `chrome.experimentalActor`:
What is the current status and long-term vision for `chrome.experimentalActor`? Is there a proposal or roadmap to eventually stabilize and expose a public `chrome.actor` Extension API for third-party extension developers? Or is this API path intended strictly for internal testing and prototype integration?
2. Architectural Rationale: WebUI (`chrome-untrusted://`) vs. Extension Path:
We noticed that Google chose to implement GLIC as an Untrusted WebUI component (`chrome-untrusted://glic/`) communicating via C++ Mojo IPC, rather than building it as a standard Chrome Extension. Given that GLIC’s Web API surface (`glic_api.ts`) encompasses several capabilities already available in standard extension APIs (such as `chrome.tabs`, `chrome.sidePanel`, and `chrome.windows`), what were the primary security or architectural trade-offs (e.g., privilege separation, prompt injection isolation, multi-platform support, or UI framing) that led to choosing the WebUI path over an extension-based architecture?
Thank you very much for your time and guidance!
Best regards,
Gnanasekar