[Talk-ca] Crowdsourcing buildings with Statistics Canada

[Stewart C. Russell]

Hi John - some great points here.

> My understanding is currently he’s looking getting hold of the City of
> Ottawa building outline data and making it available to OpenStreetMap
> without the current license restriction.

This would be wonderful. It would be ideal if the data could be placed
on data.gc.ca and use the OGL-CA v2 licence. OSM can't use any data
under the City of Ottawa Open Data - Terms of use
<http://ottawa.ca/en/mobile-apps-and-open-data/open-data-terms-use>. I
also have my doubts about the acceptability of the Statistics Canada
Open Licence Agreement <http://www.statcan.gc.ca/eng/reference/licence>.
OGL-CA v2, though, we know to be acceptable.

Also, if there were to be an import, we *must* follow the
Import/Guidelines
<https://wiki.openstreetmap.org/wiki/Import/Guidelines> or risk having
any new imports deleted. The recent LA building import provides a decent
template, but there are no imports without the Data Working Group having
knowledge of it.

[** Bjenk: if all this seems gibberish, please ping me off-list, and I'd
be happy to have a chat. Despite my previous flippant comments, I think
this is a great project.]

To some more of John's points:

> He’s also asking for the building outline to be tagged with the address
> including postcode.  Which is interesting as currently each node of
> store within a building might have part of the address.

For sure. I looked at the City of Ottawa data, and getting it to mesh
with existing address points and ranges in OSM is going to be challenging:

* fixing street naming to meet OSM standards (so Ottawa's 991 CARLING
AVE would have to become addr:housenumber=991 and addr:street=Carling
Avenue). Not impossible, but would need some manual oversight

* Inconsistent application of French to some street names, English to
others, and no obvious metadata to distinguish language

* some buildings in mixed-use neighbourhoods will have multiple address
points, all containing the same address (eg St Stephen's on Parkdale Ave
has three 579 Parkdale Ave nodes)

* some buildings just plain don't have address points nearby (like the
Agri-Food Canada Building on Carling Ave)

* rationalizing address points with existing address ranges.

And then there's the postal code problem. If Stat Canada can bring us a
licence-compatible data set of full codes that Canada Post *won't* try
to sue us over, that would be glorious. I'm not sure we could get enough
traction with the general Canadian public to do the "Free the Postcode"
initiative like in the UK to make this useful as a crowdsourcing effort.

> … One problem I see arising is a new mapper mapping to the
> Stats Canada guide lines using iD changes one or more existing tags.  I
> do a fair amount of validation in HOT and some newer mappers either
> completely ignore or misunderstand the instructions.

Yes, this can be a problem with newer mappers. There would need to be a
careful data quality metric, but also an understanding that unpaid,
crowdsourced data may always have errors.

Big project. Genuine opportunities for learning and value on all sides.

cheers,
 Stewart