The GB Roads Visualisation builds a map of A-roads and motorways in Great Britain, based on the locations of traffic watch points (part of the transport data set made available through data.gov.uk).
The map is semi-schematic in nature, as the true path of each road can only be inferred based on the watch point locations, which may be infrequent in more rural areas. As the map is being draw it reveals the shape of the numerical numbering system used for roads in Great Britain: note how the roads appear first in the east of England, revealing the characteristic coastline, followed by the south east and then the south coast.
Very straight roads shown on the map are generally not Roman roads(!), but a combination of quirks in the implementation and quirks in the road numbering. The implementation quirks reflect the fact that it's hard to know precisely where a road starts and ends and is therefore hard to "join all the dots" together correctly, hence some roads are a mix of squiggly and straight lines. The more interesting cases are those where a road number is not unique but is used in two different locations that are connected only by other roads. A good example of this is the A5095, which appears to run straight from the South Midlands to the Lake District. In fact, there are two distinct sections of road numbered A5095, one in Northampton and the other near Ambleside. This is the kind of insight which is hard to gain from looking just at raw data and may only be apparent with a visualisation.
The visualisation is implemented as a Flash movie, which retrieves the data it needs by issuing SPARQL queries to the data.gov.uk Transport SPARQL endpoint. The data points being fetched at any one time are shown across the bottom of the screen, while the various parts of the visualisation run more slowly to make them more comprehensible. Each complete rendering of the visualisation results in approximately 1800 queries to the SPARQL endpoint.