[Talk-ca] import complaints

Richard Weait richard at weait.com
Fri Dec 2 21:07:11 GMT 2011


On Fri, Dec 2, 2011 at 3:27 PM, Connors, Bernie (SNB)
<Bernie.Connors at snb.ca> wrote:
> Richard,
>
>        Do you have a link to Import Guidelines that are specific to Canvec data?

Sure.

All imports should comply with the OSM import guidelines.  My
preference is that we do not import at all. We should treat outside
sources the way we treat aerial imagery.  This is a deliberately
provocative statement.  More below.

http://wiki.openstreetmap.org/wiki/Import_guidelines

The automated edit guidelines apply to imports, and to any mass / bulk edit.

http://wiki.openstreetmap.org/wiki/Automated_Edits/Code_of_Conduct

Now, on imports.  I'm very grateful that NRCan has been generous and
allowed us to include Canadian government data in OSM.  It's even
better that folks at NRCan participate in the OSM community.  This is
not about them, and they are not at fault.

It is my opinion that imports delay or end the growth of local OSM
communities at the expense of "some data is better than perfect data".
 How?  In the old days, a person might hear of OSM, look at the map
and see that their neighbourhood is a blank spot on the map.  That
motivated them to map their neighbourhood, and "boom" all of a sudden
that new mapper is $SuperMapper.  Pick one of the old-timers on the
talk@ list for a value for $SuperMapper.  Today, a new mapper might
look at rural Ontario, and think, "Ah, all the roads are there.  No
need for me to map."  And we have missed the opportunity to create a
new SuperMapper.

So this might be true of any place, where mappers have mapped from a
distance.  Why pick on imports?  Imports are too fast and too easy.
That leads to insufficient care being applied to each node and way.

When we map from aerial imagery we carefully consider each node placed
on each way, because we have to do them all one by one, based on what
we interpret from the imagery.  That's good.  With an import, we might
look at the 20 km**2 and check a few spots, but it is not possible to
give the same attention to each and every node that we would as we
draw them by hand.

So we get broken imports because we don't pay enough attention.  Our
tools have improved over the last seven years to reduce the gross
errors that we make with imports but that is no substitute for the
individual care and attention that we give to the nodes and ways we
create through hard(er) work.

So imports are worse than referring to an external source like tracing
aerial imagery.  The quality is lower.

And the result can prevent or dissuade new mappers from joining OSM.

In the alternative universe, where we did not import, new mappers
found their neighbourhoods as blank spots and started mapping.  In
that alternative universe talk-ca has 10X or 100X readers.  Every town
in Canada has one or more local mappers.  today, we might say every
city in Canada has one or more mappers.  In ten years we might have a
mapper in every town.  In that alternate universe, ten years from now,
perhaps there is a mapper in every Canadian village.

Is there a difference?  Yes.  We want a mapper on every block.
Imagine, if a coffee shop around the corner changes name, how long
would it take to update in OSM with a mapper on every block?  "Not
long" is the answer.  With only one or more mappers in the nearest
city, OSM will never know.  We want a mapper on every block and if
imports are discouraging that, we need to fix the way that we are
doing imports.

We could ban imports.  But we still want to have access to external
sources.  So let's start treating external like we treat aerial
imagery.  When you do a foot survey you take notes and photos and draw
sketches.  Then you map it by referring to your notes and photos and
sketches and aerial imagery.  That's how we notice that the aerial
imagery is three years old doesn't show the new shopping plaza or
extension on the old mill.  And we consider all of those sources then
take the best we can from every source and put it in OSM.  That's why
OSM is so good where we have a rich community.  OSM is better than any
other single source.

On the other hand, just doing an import can only be as good as that
single source.

So let's start using external sources as one of many sources for our
mapping, not a replacement for mapping.

Best regards (and catching my breath after that rant)
Richard.



More information about the Talk-ca mailing list