[OSGeo-Discuss] Is Your Project In OSGeo Labs?

Stephen Woodbridge woodbri at swoodbridge.com
Sat Dec 8 17:10:51 PST 2012


HI Milo,

There are various initiatives related to Geocoding and it would be good 
if could share insights and resources. Some of the projects are:

http://www.pagcgeo.org/
http://www.postgis.org/                  - has Tiger based geocoder
http://wiki.osgeo.org/wiki/OpenGeocoder  - list some of these and others
http://geocoder.us/
http://www.opengeocoder.net/             - some Microsoft employee
openstreetmap nominatim

The big differences between the various projects are based primarily on 
three dimensions:

1. Sources data ie: Tiger only, OpenStreetMap data, any data
2. Language and Geocoding Strategy
3. Licensing

Combining/collabation of efforts really depends a lot of the 
compatibility of these dimensions.

Package   DataSource(s)     Language    Licensing
-----------------------------------------------
pagc        Multiple           C          MIT-X
             Street segments
             parcels
             landmarks
------------------------------------------------
postgis     Tiger             pgpsql      GPL2
------------------------------------------------
geocoder.us Tiger              Perl       Perl
------------------------------------------------
nominatim   OSM data            ??        ODbl?
-------------------------------------------------

Pagc, postgis geocoder and geocoder.us all started out as Tiger data 
based geocoder and have expanded into other realms that use similar 
structured data. These projects have more in common with one another 
than any of them have with nominatim, unless I am terribly mistaken. 
That is not to say we should not collaborate only the data structures 
and strategies for querying them are very different.

For example, I just wrote a custom geocoder taking a small amount of 
code from pagc and wrapping it into a pgpsql stored procedure and then 
wrote a query planner in pgsql to do geocoding. I was able to load 50M 
records derived from Tiger, and index them in 6 hrs on a slow 4GB linux 
box and can geocode a table 216K addresses in about 44 ms per record. 
This is conceptually similar to the processes imployed by the postgis 
geocoder and by geocoder.us and for that matter a 2-3 other geocoders 
that I have worked with in the past that were not open source, one of 
them being a geocoder I wrote that only worked with the old Tiger/Line data.

I have looked at nominatim, a couple of times to try an understand the 
process, but it seems to be very tied to the OSM data structures and 
infrastructure. So if someone wants to setup a private nominatim service 
for a client, they basic need LOTS of iron, LOTS of disk, and need to 
mirror much/most?/all? of the OSM infrastructure. And the licensing 
needs to be approved by the client.

If I want to create a geocoder that works with Navteq data for a client 
that has access to that data, it is pretty easy for me to load the data 
into PAGC, or the geocoder I just wrote, or even the postgis or 
geocoder.us code with a little massaging of the data. I'm not sure where 
I would start with the nominatim code.

OK, I admit that the failure here is my knowledge about Nominatim and/or 
maybe that Nominatim/OpenStreetMap are not interested in solving the use 
cases that I presented.

Sorry, I have gotten off on a rant. Back to your point about joining 
forces, I'm ok with, but I'm not sure where to start. I think I outlined 
the problem as I see it, but some additional information related to 
Nominatim could help me get beyond my biases if I can see how we can 
successfully collaborate.

Best regards,
   -Steve

On 12/8/2012 5:19 AM, Milo van der Linden wrote:
> Would it be good if opengeocoder joins forces with openstreetmap nominatim?
>
> Op 29 nov. 2012 03:05 schreef "Stephen Woodbridge"
> <woodbri at swoodbridge.com <mailto:woodbri at swoodbridge.com>> het volgende:
>
>     On 11/28/2012 7:31 PM, Landon Blake wrote:
>
>         I'm in the process of trying to take over as the steward for OSGeo
>         Labs as part of my duties with the OSGeo Incubation Committee.
>         As part
>         of this process I'd like to get a handle on the projects that
>         are "in"
>         labs. There is a short list of "stable" and "young and experimental"
>         projects on the current Labs wiki page. Since I'm editing that page
>         today, here is the list:
>
>         Stable Projects:
>         - GeoWebCache
>         - pgRouting
>
>         Young and Experimental Projects
>         - GeoExt
>         - GeoFunctions
>         - Geoinformatica
>
>
>     I think these are more or less mine:
>
>         - OpenGeocoder
>         - OpenRouter
>
>
>     There is an OpenGeocoderRouter list that I started but there is no
>     viable activity on it at this time.
>
>     OpenRouter is a project related to internet routing. I started
>     OpenGraphRouter using a GSoC project to get started. The goal was to
>     create a routing solution that was MIT-X licensed instead of GPL. We
>     have sine joined forces with pgRouting and are developing the code
>     that is MIT-X algorithms, which can be bundled with pgRouting
>     effectively making them dual licensed.
>
>     OpenGeocoder.net appears to be Steve Coast @ Microsoft and not
>     related to OSGeo stuff.
>
>     http://wiki.osgeo.org/wiki/__OpenGeocoder
>     <http://wiki.osgeo.org/wiki/OpenGeocoder> never got off the ground,
>     but I have been working with PAGC over the last few years. Our big
>     issue at the moment is addressing some serious performance issues
>     when you scale up from county level data sets to national data sets.
>     Basically it is just me and Walter, the developer, working behind
>     the scenes on these technical issues. Once these are resolved I hope
>     to see if we can some activity going again with this.
>
>     On a side note, I have take the address standardizer from PAGC,
>     built it as a library and wrapped it into a postgresql stored
>     procedure extension. Based on that I have prototyped up a Tiger
>     geocoder that works very well and is very fast. I'm still work on
>     various things so it is not ready for prime time but this might
>     eventually become OpenSource also.
>
>     I'm not sure what it means or how you get a project like these "in
>     labs" but these are mostly orphaned except I have an interest in
>     them and will respond to queries about them.
>
>     Thanks,
>        -Steve
>
>         - Grids
>         - OSGeo Graphics
>         - pycsw
>         - OWSLib
>         - SemanticGeo
>         - ZOO-Project
>
>         Can you please let me know if you are involved with one of these
>         projects? I'm trying to determine which projects are "in" labs, and
>         then establish a point of contact with each project so I can
>         help them
>         get ready for official incubation.
>
>         Thanks.
>
>         Landon
>
>         P.S. - If you have thoughts on the purpose and work of OSGeo Labs,
>         please let me know. I have my own vision, but I'd like to get
>         feedback
>         from other OSGeo members.
>         _________________________________________________
>         Discuss mailing list
>         Discuss at lists.osgeo.org <mailto:Discuss at lists.osgeo.org>
>         http://lists.osgeo.org/__mailman/listinfo/discuss
>         <http://lists.osgeo.org/mailman/listinfo/discuss>
>
>
>     _________________________________________________
>     Discuss mailing list
>     Discuss at lists.osgeo.org <mailto:Discuss at lists.osgeo.org>
>     http://lists.osgeo.org/__mailman/listinfo/discuss
>     <http://lists.osgeo.org/mailman/listinfo/discuss>
>




More information about the Discuss mailing list