Firstly, cheap (and therefore pervasive) GPS devices are built to a cost, and the aspect of quality that suffers is usually satellite lock and/or power efficiency. Mine is great for geotagging photos in the Dales, but at best so-so in urban environments. There are several locations I have tried in Manchester and London where it simply will acquire not enough satellites and it is therefore no more accurate than asking a policeman. (Or scally, if truly desperate. If you see a cockney in what you thought was Manchester then I suppose you know you are REALLY lost…)

Secondly, this highlights how important user-generated content is, because the exact location has frequently been posted to Google by the great unwashed (hence the “unverified” tag.)

On the “thieving bastards” example, ISTR there was a story last year about a couple of Northern Rail employees receiving actionable abuse on Facebook, will see if I can dig it out.

On an unrelated note, what the hell happened to numbering buildings on streets? On Whitworth St in Manchester for example, I would say about every tenth building is numbered and the rest are named or have only company information on the facade. Makes for a lot of unnecessary walking up and down to home in on the last 50 yards.

“Computer algorithms giving directions will lead people alongside the road (or maybe, if they are smart enough, a footpath). ”

The algorithms will do what they have been asked to do with the data that they have – so we need to switch from algorithms to heuristics and other mining techniques; and as you say, we need better granularity in the data. We also need some common sense in the planning system to spot dumb questions. The response to “when is the next Tube to Charing Cross from Embankment” should be “I suggest you walk instead.”