Add proposal for runtime.onEnabled and runtime.onExtensionLoaded events

[sohailrajdev97]

Creating a proposal for #353. This is based on @justinlulejian's proposal document here:

https://docs.google.com/document/d/1Fpqg-cQRoNkQVL-hoQi5fqhVoOEdjOiGoQXWdw_yFxo/edit?tab=t.0

[hanguokai]

Thanks for writing the formal proposal and sponsoring the implementation.

[oliverdunk]

Thanks so much for picking this up! I left one comment, but otherwise LGTM.

[xeenon]

See above comments.

[sohailrajdev97]

@Rob--W - Can you also take a look at this one?

[oliverdunk]

LGTM - thanks! Seems like the discussion on enum values is resolved.

[Rob--W]

I reviewed the proposal, and it generally looks good. There is one unspecified aspect... What are the expected semantics when incognito: "split" is used? This is currently only implemented by Chromium browsers (e.g. Firefox does not support it).

When "incognito": "split" is specified in manifest.json, a new extension instance (background script) starts when the first Incognito window opens. When the last incognito window closes, the incognito session including extensions are torn down. In this model, my expectation is that Chrome may consider firing onExtensionLoaded whenever a new incognito window is opened, even if a non-incognito window has already been initialized.

The proposal currently suggests to use the onExtensionLoaded event to decide on whether to populate storage.session. That use case does not require a new event: an extension could unconditionally call storage.session.get at startup and check whether the value was already populated, and if not, start the initialization. Note that in Chrome, storage.session in incognito and non-incognito modes are separate.

[oliverdunk]

I reviewed the proposal, and it generally looks good. There is one unspecified aspect... What are the expected semantics when incognito: "split" is used?

This is a great callout. I think https://issues.chromium.org/41029550 which you opened is good context as well. This is an existing problem with the onInstalled event and it would be nice to agree on something as part of this work. My first thought is that firing the event probably makes sense but I'd be interested to hear any other arguments.

The proposal currently suggests to use the onExtensionLoaded event to decide on whether to populate storage.session. That use case does not require a new event: an extension could unconditionally call storage.session.get at startup and check whether the value was already populated

I think we can probably just remove that example from the proposal. The other use cases are compelling and harder to do with existing workarounds.

[sohailrajdev97]

This is an existing problem with the onInstalled event and it would be nice to agree on something as part of this work. My first thought is that firing the event probably makes sense but I'd be interested to hear any other arguments.

I too agree that firing the event for split mode incognito workers makes sense. I have updated the proposal accordingly. Please take a look 😄.

FYI @Rob--W, @oliverdunk, @rdcronin, @xeenon

[oliverdunk]

Looks good to me. I confirmed that Devlin was happy with the split mode behavior from a Chrome perspective.