[OSM-talk] House numbers... One more suggestion

Jochen Topf jochen at remote.org
Tue Jul 29 07:43:39 BST 2008


Maybe some words about how the whole house number stuff in Karlsruhe
came about, will help understand, what we had in mind:

We looked at the different proposals and saw that there was a difficult
problem that had gotten stuck. Many ideas, but no progress. But it needed
to be addressed, because addresses are so important for routing and other
uses of the data.

So we organized a one-day workshop in Karlsruhe with a few people with
the specific goal of getting one step forward on this problem. Not to
solve all problems (that would have taken more than a day), but to
identify what we could do, to get the thing unstuck. Some of us
collected some data in the days before so that we had samples.

We were looking for a solution that would
* allow Joe Average Mapper to contribute data, so it needed to be
  conceptually simple
* would work in Germany (bonus for solutions that work in other places,
  too)
* not require huge amounts of new software to be written (because then
  we would have been stuck again, because nobody had time to write
  it)
* make common cases easy and complex cases possible
* be robust in the face of common operations like splitting ways

We were explicitly NOT looking for a solution that would solve the
problem for the whole world/for everybody/for all times. Look at the
OASIS xAL standard (http://www.oasis-open.org/committees/ciq/ciq.html#6)
if you want to see the monster such as solution would require.

We looked at the proposals in the Wiki and had some ideas of our own.
They all failed on one or more of the above points. After discussing the
whole thing for a few hours we developed the solution that is now known
as Karlsruhe schema. It has many drawbacks, but it is reasonably good on
those points listed above.

Unlike the other proposals it can actually be used today, there was some
sample data and a rendering. Thats why it was picked up. I am a computer
science person myself and from a data modelling perspective I cringe
when I see what we have produced, but we can not deny that it works (at
least sort-of). We do not think of this as the end of the discussion,
but as one step in the discussion.

The Karlsruhe schema (and addressing in OSM in general) still has two
huge hurdles in front of it: a) when and how will mappers actually
contribute data in larger quantities (there are not even 5000 house
numbers in the whole of Europe mapped currently). b) how can this data
be used in real applications. There certainly is a lot of work to be
done. At the workshop in Karlsruhe we actually created a piece of
software that could search for theses addresses, but it was so clumsy
and hard to set up that it didn't live beyond the workshop day. And as
some critics pointed out there is some complex programming to do, before
the data can be used for routing or other things.

I hope this has shed some light on the Karlsruhe schema. I invite
everybody to work on the Karlsruhe schema to improve it, to collect data
and to create software for it. Or, if you prefer, to work on other
schemes, collect data and create software. Once we have some experience
with the different options, it will be easier to see, which one works
best.

Jochen
-- 
Jochen Topf  jochen at remote.org  http://www.remote.org/jochen/  +49-721-388298





More information about the talk mailing list