A few of us are working on a dial test app for the presidential debate. During the debate you will continuously adjust a slider based on your favorable or unfavorable response at any given moment. On the screen you will see an aggregate line graph of the current sentiment of undecided voters. CNN does this with about 40 people, which you can see here: http://bit.ly/RQVGv2 We want to open that up to the entire country.<p>What technical challenges are we going to face? What is the best way to collect and display this data in realtime at scale (e.g. long polling, websockets)? If you have experience in this area we would appreciate a heads up with anything you've learned. If you have experience in this area we would appreciate a heads up with anything you've learned.<p>We are trying to optimize for two scenarios:<p>1. Take in and send out data during the debate to provide an engaging real-time experience<p>2. Provide a complete dataset for analysis after the debate. The resolution on this data will likely be greater than the resolution in scenario 1.
I built something like this for a hackday (actually it's still up <a href="http://worm.cloudfoundry.com/" rel="nofollow">http://worm.cloudfoundry.com/</a>) - I only ran it with 50 people so I can't really speak about the scale, but I one thing I found was that lag can be a significant issue.<p>Different people will have different lags so you need to take that into account and adjust the historical view appropriately (i.e you can't just chuck all incoming data into a "current" bucket).