Someone's probably already thought of this, but... Presumably it would be possible to programmatically generate all possible postcodes. The space wouldn't be all that large, because we know all the prefixes now as freethepostcode and npemap.org.uk have collected them.
Then, in a similar fashion to the NameFinder Postcode Lookup: http://www.frankieandshadow.com/osm/ you could Google for each one, and parse the resulting pages for something looking like an address. Then, for each hit where you found one, in Mechanical Turk fashion, you could present willing volunteers with: - The generated postcode - The possible address text you found - The OpenStreetMap map for the area and tell them that if it does look like a valid address, and if they can see the named road, to click on it - or, if not, click "Reject". You would then store that as a location for that postcode, marked with something which says "this isn't completely verified, and may be a bit vague, but it's a lot better than nothing". I think you could do about 300 of these an hour (1 every 10 seconds or so). This would generate a reasonably-accurate postcode database with much wider coverage fairly quickly. Given that each postcode would be taken from a different web page, there's a good chance there'd be no encumbering legal issues. Of course, the postcode data may well be CC-BY-SA. So it doesn't obviate the need for freethepostcode and npemap. But it would quickly get OSM to a stage where it could do fairly comprehensive postcode lookup. Another, better, source of addresses for this process would be mailing lists borrowed from sympathetic mail order companies. Clearly, they could only let you have addresses they had permission to give you, but that might be a fair few. Gerv _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk