<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 5.5.2657.73">
<TITLE>Re: [OSM-talk] postcode idea</TITLE>
</HEAD>
<BODY>

<P><FONT SIZE=2>Hi All,</FONT>
</P>

<P><FONT SIZE=2>It turns out that you don't want the street number of every house and  </FONT>
<BR><FONT SIZE=2>business in an area (which diminishes the privacy concerns). Instead  </FONT>
<BR><FONT SIZE=2>you want a more limited set of information. A standard road network  </FONT>
<BR><FONT SIZE=2>file contains (at least in Canada and the US, which is what I know as  </FONT>
<BR><FONT SIZE=2>an American who is a Canadian resident) for each road segment:</FONT>
</P>

<P><FONT SIZE=2>1. The street name.</FONT>
<BR><FONT SIZE=2>2. The street "direction" (N, S, W, E, NE, SW, etc.) for streets like  </FONT>
<BR><FONT SIZE=2>West 10th Ave or Deer Trail SE. This value is missing for streets  </FONT>
<BR><FONT SIZE=2>that do not have a "direction".</FONT>
<BR><FONT SIZE=2>3. Address range information that consists of (a) the starting number  </FONT>
<BR><FONT SIZE=2>on the left-side of the street, (b) the starting number on the right- </FONT>
<BR><FONT SIZE=2>side of the street, (c) the ending number on the left-side of the  </FONT>
<BR><FONT SIZE=2>street, and (d) the ending number on the right-side of the street.  </FONT>
<BR><FONT SIZE=2>These are often called Left and Right From-To pairs.</FONT>
<BR><FONT SIZE=2>4. Local geographic area identifiers. In US TIGER/Line data this is  </FONT>
<BR><FONT SIZE=2>the five digit zip code. In the UK the likely equivalent would be the  </FONT>
<BR><FONT SIZE=2>first block of characters (e.g., NN1, or AB32) of the postcode. In  </FONT>
<BR><FONT SIZE=2>Canada the ideal combination is the first three characters of the  </FONT>
<BR><FONT SIZE=2>postal code _and_ the "city". Both are needed in Canada for handling  </FONT>
<BR><FONT SIZE=2>sparsely populated rural areas, which we have a lot more of than is  </FONT>
<BR><FONT SIZE=2>the case in the UK ;-). The local geographic identifiers can be  </FONT>
<BR><FONT SIZE=2>different for one side of the street versus the other. The dividing  </FONT>
<BR><FONT SIZE=2>line between Vancouver and Burnaby, British Columbia is a street  </FONT>
<BR><FONT SIZE=2>called Boundary Road. In this case, both the city and first three  </FONT>
<BR><FONT SIZE=2>characters of the postal code depend on whether you are talking about  </FONT>
<BR><FONT SIZE=2>a location on the west or east side of the street.</FONT>
</P>

<P><FONT SIZE=2>With this set of information one minimizes privacy concerns since the  </FONT>
<BR><FONT SIZE=2>address information is very incomplete, but there is enough  </FONT>
<BR><FONT SIZE=2>information to do things like geocode a particular street address  </FONT>
<BR><FONT SIZE=2>with a reasonably high degree of accuracy (not sub-metre, but not  </FONT>
<BR><FONT SIZE=2>bad). Other bits of information that would be nice additions to each  </FONT>
<BR><FONT SIZE=2>road segment are:</FONT>
</P>

<P><FONT SIZE=2>1. Whether the street is one-way, and the direction of travel on that  </FONT>
<BR><FONT SIZE=2>street if it is one way.</FONT>
<BR><FONT SIZE=2>2. The posted speed limit on the road.</FONT>
<BR><FONT SIZE=2>3. The number of lanes of traffic.</FONT>
<BR><FONT SIZE=2>4. Whether the road has a bike lane.</FONT>
<BR><FONT SIZE=2>5. Whether the road has a sidewalk.</FONT>
</P>

<P><FONT SIZE=2> From this additional information the road map can then be used for  </FONT>
<BR><FONT SIZE=2>routing/road directions (customized for walkers, bikers, or drivers)  </FONT>
<BR><FONT SIZE=2>and constructing drive-time, walk-time, bike-time polygons.</FONT>
</P>

<P><FONT SIZE=2>How would I collect this data? For postal codes I would do exactly as  </FONT>
<BR><FONT SIZE=2>proposed. In Canada and the US people don't generally mind telling  </FONT>
<BR><FONT SIZE=2>you their postal/zip codes. I would include an explanation of why you  </FONT>
<BR><FONT SIZE=2>are doing this, and an indicate to people that the data will treated  </FONT>
<BR><FONT SIZE=2>in a way that will protect their privacy, indicating the way that  </FONT>
<BR><FONT SIZE=2>this will be done. One thing you may run into (based on what I know  </FONT>
<BR><FONT SIZE=2>from CA/US) is that in urban areas a particular block may have  </FONT>
<BR><FONT SIZE=2>multiple postal codes. I've already mentioned what can happen on  </FONT>
<BR><FONT SIZE=2>different sides of a street with respect to postal codes. It turns  </FONT>
<BR><FONT SIZE=2>out that heavy mail recipients may have their own postcode that is  </FONT>
<BR><FONT SIZE=2>different from other locations in the same block, on the same side of  </FONT>
<BR><FONT SIZE=2>the street. For instance, in residential areas with mixed single  </FONT>
<BR><FONT SIZE=2>family houses and apartment buildings, an apartment building may have  </FONT>
<BR><FONT SIZE=2>one postal code, but the single family houses on either side of it  </FONT>
<BR><FONT SIZE=2>may have a different postal code that is common to both houses. In  </FONT>
<BR><FONT SIZE=2>terms of the basic road segments, that is what OSM already does so  </FONT>
<BR><FONT SIZE=2>well. The major change I would make is increasing the emphasis on  </FONT>
<BR><FONT SIZE=2>street intersections. I would specifically geocode the location  </FONT>
<BR><FONT SIZE=2>(possibly from all corners) of an intersection and then note the  </FONT>
<BR><FONT SIZE=2>street name, street direction, street number, the one-way road  </FONT>
<BR><FONT SIZE=2>information, the sidewalk/bike lane information at each corner. It  </FONT>
<BR><FONT SIZE=2>seems natural to me to define a road segment as running from one  </FONT>
<BR><FONT SIZE=2>intersection to the next intersection. If it is done this way, then  </FONT>
<BR><FONT SIZE=2>it will be pretty easy to attach the left and right from-to pairs of  </FONT>
<BR><FONT SIZE=2>street numbers, street names, street directions, and one-way road  </FONT>
<BR><FONT SIZE=2>indicators to the road segments. It turns out that the intersection  </FONT>
<BR><FONT SIZE=2>data is valuable in-and-of-itself since it allows for easy geocoding  </FONT>
<BR><FONT SIZE=2>of intersections as opposed to addresses. I might know that a store  </FONT>
<BR><FONT SIZE=2>is near the corner of 4th Ave and Main St, but don't have a clue as  </FONT>
<BR><FONT SIZE=2>to its exact address.</FONT>
</P>

<P><FONT SIZE=2>While the above is based on my CA/US experience, my guess is that  </FONT>
<BR><FONT SIZE=2>nearly all of it will be applicable for the UK as well (I'm not sure  </FONT>
<BR><FONT SIZE=2>if the street "direction" information is applicable). Having read  </FONT>
<BR><FONT SIZE=2>over some of the Royal Post's documents, there is a lot of similarity  </FONT>
<BR><FONT SIZE=2>with how things are done in all three countries. A full Canadian  </FONT>
<BR><FONT SIZE=2>postal code, UK postal code, and US zip+4 code get down to the block- </FONT>
<BR><FONT SIZE=2>face level (e.g., one block on one side of the street) in most cases  </FONT>
<BR><FONT SIZE=2>in all three countries.</FONT>
</P>

<P><FONT SIZE=2>Sorry for the length of this post.</FONT>
</P>

<P><FONT SIZE=2>Dan</FONT>
</P>

<P><FONT SIZE=2>On 30-Mar-06, at 1:55 AM, Nick Hill wrote:</FONT>
</P>

<P><FONT SIZE=2>> Immanuel.scholz@gmx.de wrote:</FONT>
<BR><FONT SIZE=2>>> I don't know whether all of these people agree to have their  </FONT>
<BR><FONT SIZE=2>>> address listed in an available online database (regardless of the  </FONT>
<BR><FONT SIZE=2>>> fact that their names are removed, which could be easily obtained  </FONT>
<BR><FONT SIZE=2>>> from other sources).</FONT>
<BR><FONT SIZE=2>></FONT>
<BR><FONT SIZE=2>> Ultimately, if we are building an on-line database with road names,  </FONT>
<BR><FONT SIZE=2>> locality names, postcodes and street numbering, all addresses will  </FONT>
<BR><FONT SIZE=2>> appear on it.</FONT>
<BR><FONT SIZE=2>></FONT>
<BR><FONT SIZE=2>> If we accept this, then whether we use an address to help structure  </FONT>
<BR><FONT SIZE=2>> the system perhaps becomes practically neutral.</FONT>
<BR><FONT SIZE=2>></FONT>
<BR><FONT SIZE=2>> We then have the point of law Copyright: Does a singular address  </FONT>
<BR><FONT SIZE=2>> provided to someone constitute a copyrightable work? If so, when I  </FONT>
<BR><FONT SIZE=2>> have that address, who owns the copyright?</FONT>
<BR><FONT SIZE=2>></FONT>
<BR><FONT SIZE=2>></FONT>
<BR><FONT SIZE=2>> _______________________________________________</FONT>
<BR><FONT SIZE=2>> talk mailing list</FONT>
<BR><FONT SIZE=2>> talk@openstreetmap.org</FONT>
<BR><FONT SIZE=2>> <A HREF="http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk" TARGET="_blank">http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk</A></FONT>
</P>

</BODY>
</HTML>