Followup: modularity, WebExtensions, and going faster

[Andrew McKay]

On 12 October 2016 at 12:57, Ian Bicking <ibicking at mozilla.com> wrote:
> I also think exactly the kind of pacing of these ideas makes WebExtensions a
> hard fit, though Hybrid WebExtensions would be a good fit.  Page Shot could
> be rewritten fairly easily to be 95% WebExtension.  The last 5% could be
> fixed with probably-acceptable product changes, but we'd be left
> wondering... is there something we could do outside the boundaries of
> WebExtensions that would be an important and valuable product direction? The
> data-driven process we want to do requires us to actually ship those ideas.
> And then be ready to un-ship them as well, ideally without making the
> WebExtension team run around for our capricious and unfounded requests.

I'm thinking it would be good to make the main Test Pilot a
WebExtension Experiment [1]. When a new project comes into Test Pilot,
if the WebExtensions APIs don't exist then they get added to the main
Test Pilot add-on as WebExtension APIs [2].

In this case you could possibly write that remaining 5% as a
WebExtension API (or not, the choice is yours). If then Page Shot
graduates, we could graduate the API, or make whatever decision we'd
like at that time.

This way we get to use the same data pipeline and use that to guide the APIs.

I've been wanting to post more about WebExtension Experiments but
wanted to get the documentation and flow more in order, so I know
there's still some work to be done there.

[1] https://webextensions-experiments.readthedocs.io/en/latest/
[2] This would require us to allow Test Pilot as a WebExtension
Experiment to exist on release.