Stand Up Voice Fill Servers

[johngruen]

In order to migrate Voice Fill to a server-based implementation, we will need dev, staging, and production environments.

[ghost]

This'll take some organizing. A rough list:

• [ ] What's our timeline for this?

• [ ] Scalability plan. What are the chokepoints and how are we addressing them? How many users are we expecting?

• [ ] Deployment plan. How do we do deploys? (This is pretty simple if it's only one server)

• [ ] Monitoring plan. What are we monitoring aside from general server health. If this is just a webpage, we're good. If it's other services, we should identify those points.

• [ ] Metrics plan. How does this change what we're collecting?

• [ ] QA plan. We'll need to update that to include any changes here

• [ ] add monitoring end points to code (like /__version__. I can get a list of these, they are simple)

• [ ] security review. We'll need to update andre's flow diagram on how the system works and get it reviewed

• [ ] (are we collecting any user data? If so...) legal and privacy reviews

[cascella]

I hope you will let the user choose if they want to use the server-based voice recognition or prefer to keep using the local one. Ideally, the choice would be offered explicitly when the extension is installed. If Mozilla wants to stand for user privacy and it is important to be upfront about this.