Signature Fallback

[thomasdavis]

Great work so far, I think we should operate under the assumption that We The People might actually go down.

Not sure if this server supports that, but I imagine some sort of queuing setup, where signatures are saved there first and foremost and sent off by a worker.

So if We The People goes down, we can at least send off those signature requests later.

The rough design is that the API drops requests to a Redis instance.

There'll be a backend daemon that will then read from that Redis instance and make the API calls to the WtP API. That's mostly required as they rate limit to ~1.5K requests per minute and Sina estimated peak traffic spikes well in excess of that.

I don't know the performance characteristics of Heroku very well, but it's a very simple node service, so should be able to deal with decent QPS.