Native Messaging (Was: Core API Proposal)

> On May 4, 2016, at 03:56, Andrew Swan <aswan@mozilla.com> wrote:
> 
> 
> On Mon, May 2, 2016 at 10:29 PM, Florian Rivoal <florian@rivoal.net <mailto:florian@rivoal.net>> wrote:
> Communication with native processes is indeed an important feature, and I agree that we should look at, but not necessarily copy, what google has been doing in this area, and compare that with alternative approaches.
> 
> I suspect that the first iteration of a specification for extensions might skip this feature, and first focus on standardizing the core parts before turning to this kind of more complex (and probably more controversial, since there are several possible approaches) features.
> 
> However, if people want to work on it from early on, I am certainly not going to stand in the way.
> 
> It also depends on when browser vendors expect to be shipping this. The sooner they do, the earlier we should work on it.
> 
> What do others think? Is this a topic to be explored ASAP, or to be deferred until we have stabilized the basics?
> 
> I agree that it seems unproductive to try to standardize this before the building blocks that Mike outlined in the original email are in place (kudos to the whole group for getting this process going btw).
> However, we are actively working on native messaging for Firefox right now.  The driving force is that many extension developers use NPAPI plugins to spawn and communicate with native processes, and with the upcoming removal of NPAPI we need to provide them an alternative.
> The timing is unfortunate, and I'd like to bring this back to this group eventually so we can talk about an inter-operable solution, but if we don't deliver a solution soon, we'll be stranding a bunch of our existing developers.
> If you're interested in our efforts, you can follow along at https://bugzilla.mozilla.org/show_bug.cgi?id=1190682 <https://bugzilla.mozilla.org/show_bug.cgi?id=1190682>


If you're working on this now, I think it would be beneficial to have discussions in this CG as well. At the very least, documenting where you diverge from chrome and why seems useful information for later, and other members may have constructive proposals to solve the issues you run into.

It's not going to be practical to have a *stable* spec for that before the core spec is in place, but at the same time, since we're basing it off Chrome's model, we have a decent idea of where this is going. I think it could make sense to have a stub spec for native messaging that at first merely consists of a link to the (unfinished) core spec and a handwavy "do native messaging sort of like chrome" prescription, then progressively gets filled with issues about what's unclear or what's wrong with that approach, and then slowly gets fleshed out with actual spec text.

Do you think that makes sense? Would someone at Mozilla be interested in maintaining a document along these lines in this CG?

 - Florian
Co-chair