* Proxy app wrapping a request with API credentials. When requesting the config file we get the file dynamically stating which URL to use to make requests. API keys are scrubbed from the JSON file. When someone makes a request to us, wrap their request (only GET requests to specific endpoints!) and pipe the response back.
* All this works as a Connect Middleware.
* Also, make this cache responses for a given amount of time (poll time). This way people concerned with a big amount of requests can share the same resource.
1. Get [milestones](http://developer.github.com/v3/issues/milestones/#list-milestones-for-a-repository) and determine which one is ending the soonest.
1. For this milestone get both [open & closed](http://developer.github.com/v3/issues/#list-issues-for-a-repository) issues (can span [multiple pages](http://developer.github.com/v3/#pagination)).
1. Filter out issues not matching our pattern. For those that do keep tally and insert them to a map of days. Keep track of issue ids of open and closed issues.
1. Determine what the average velocity per day needs to be.
1. Go from the front (milestone creation date) to the back (milestone due date) day by day.
1. For each day that has an entry in the map, add them to the end array (for actual).
1. For expected just keep reducing the total by velocity every day.
1. Profit.
###Poll
An issue can be re-opened so we need to keep track of changes to individual tickets. Assume that polls happen quite frequently in the day and not say once a week or something.
1. Get both open & closed tickets sorted by their `created` and `updated` in a descending order. Do not need to get all pages back to UNIX time...
1. If we have a mismatch between our previous arrays of open/closed issue ids then determine if we need to move an issue (change of state) from one group to another.
