[OSM-dev] TIGER -> OSM import
tom at tom-carden.co.uk
Thu Jun 7 19:24:51 BST 2007
> Brandon Martin-Anderson wrote:
> >I know this isn't the first time this has been attempted, and it would be
> >foolish to rush in without knowing the problems encountered in the past. So
> >I'm curious: what failed in the past?
I've spoke with Steve and Mikel about this recently, and I know other
people are thinking about it and working on it too. Better wait til
Steve recovers from his jetlag for the best info.
That said, I think the main strength in OSM is the people, and the
people were not involved in the import of TIGER in the past. So,
aside from reliability issues (of both the OSM API and the machine
that was uploading it) it failed because nobody was watching it.
TIGER isn't that great a data format - e.g. roads that intersect do
not have a node at the intersection, etc. - things I'm sure you're
familiar with if you've done a routing engine that can read TIGER
data. Also TIGER doesn't care much about subtle bends in roads, and
the metadata is somewhat esoteric. Again, I'm sure you're familiar
Consensus among people I've spoken to about importing TIGER is that it
should be done piece-meal (maybe one county at a time) by people who
want it in their area. It should also be done after people have
looked at the TIGER documentation and come up with a reasonable
mapping from TIGER metadata to commonly used OSM styles (otherwise it
will have to marked up by hand, which is what we're trying to avoid).
The *ideal* scenario would be a way to grab one county's worth of .osm
file converted from TIGER, load it into JOSM, manually correct any
obvious errors or conflicts with existing data and upload it from
there. This means OSM can scale gradually to accommodate the growth
rather than struggle with millions of nodes being added all at once.
First step would be to create a wiki page with the TIGER metadata
classes and get folks to help fill in the best OSM tags to use. The
tutorials and style files for rendering TIGER with Mapserver etc will
probably help, as well as the current OSM mapnik style file.
Personally, I think the TIGER import should also add tags for all the
metadata in TIGER (possibly name-spaced/machine-tag style
tiger:foo=bar?) as well as tags about what version was imported, but I
don't think it's totally essential.
Hope that helps,
More information about the dev