I have a project based on the Global Historical Climatology Network data, going back 300 years. This is based upon readings from actual thermometers on ground based weather stations, rather than tree rings or ice cores which are far more difficult to decypher.
The thing is that when you look at the data at the lowest level (individual stations) what you're getting is small snippets of temperature information. Some stations have been around for as long as 100 years, but I don't think that there are any single stations which span the full 300 years.
http://code.google.com/p/tempgraph/
You can graph the data in various ways, all the way down to individual weather stations.