CycleStreets for example use a different database for the postcodes (Code Point Open) and if any search query looks like a post code the look it up there first. Similarly for station names they look them up in their own table first before going to nominatim.
Shaun On 31 Oct 2012, at 12:30, "Andy Robinson" <ajrli...@gmail.com> wrote: > Ah, but we are not in the business of adding non-physical stuff just because > it makes a search work better. If we have something to add postcode data to > then that's right and proper, otherwise the postcode centroid database can > be off map and referenced from a separate database. > > Cheers > Andy > >> -----Original Message----- >> From: Colin Smale [mailto:colin.sm...@xs4all.nl] >> Sent: 31 October 2012 12:26 >> To: talk-gb@openstreetmap.org >> Subject: Re: [Talk-GB] Office of National Statistics data >> >>> Colin, it's fine to add the postcode data to a node object (eg a poi), >>> just don't create a node with just the postcode and nothing else as >>> this is meaningless. The postcode data we have is not the information >>> for individual buildings, its just the centroid of the address polygon >>> which will contain any number of buildings (or post delivery points). >>> Thus when using the data its still necessary to do some interpretation >>> and it's not possible always to know that you are assigning the right >>> postcode to the right building/delivery point because we don't know >>> where the boundary of one postcode is against another for the same >>> street etc. >> >> I am guessing that just having the centroid is plenty adequate for a lot > of >> reverse geocoding probably including routing, i.e. "where is XX1 3AB" >> or "take me to XX1 3AB". Obviously it won't cover questions of the form of >> "what's the postcode for this building" which will require every > individual >> building/delivery point to be tagged. >> >> As (legal stuff permitting) "importing" the centroids would cover a very >> popular use case with (IMHO) a quality which is adequate for most people, > I >> would not be so quick to dismiss it. >> >> Colin >> >> >> _______________________________________________ >> Talk-GB mailing list >> Talk-GB@openstreetmap.org >> http://lists.openstreetmap.org/listinfo/talk-gb >> >> ----- >> No virus found in this message. >> Checked by AVG - www.avg.com >> Version: 2013.0.2742 / Virus Database: 2617/5864 - Release Date: 10/30/12 > > > _______________________________________________ > Talk-GB mailing list > Talk-GB@openstreetmap.org > http://lists.openstreetmap.org/listinfo/talk-gb _______________________________________________ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb
