In mid 90s I worked for a company that developed combined GPS/GLONASS receiver and aimed specifically at high-precision measurement market. With clever math modeling and by letting a receiver rest on one spot for longer periods of time, e.g. 30 min, they could get the precision down to 3 mm (millimeters). And that's without A-GPS, purely through oversampling of satellite data.
It's funny you mention that, as I swear the professor said it could be done to accuracy of mm, but after reviewing the wikipedia articles I had concluded I misremembered. Perhaps not...
