[Talk-GB] Separation of sources
Martin - CycleStreets
list-osm-talk-gb at cyclestreets.net
Tue Apr 6 01:35:42 BST 2010
If I may offer the view of a user of OSM data rather than a creator.
Patchy areas of OSM cause real difficulties for CycleStreets, not only in
reputational terms but in terms of debugging. Places where the basic road
pattern is absent undermine our ability to demonstrate that a UK-wide
system using open data works everywhere - nor can we even promote the
system effectively because of this. We get plenty of bug reports that
clearly arise simply as a result of missing basic road data, never mind the
subtleties of additional cycle infrastructure (which in other areas of OSM,
are stunningly good).
I'm not sure I quite understand the objection to tracing over other data
that OSM can legally and ethically use (though can appreciate David Earl's
point of view). The fact is that this OS data is "our" data which mostly we
as taxpayers have paid for already. Why shouldn't we make use of it and
incorporate it as the basis of a road network that can then be
incrementally improved, adjusted and built upon, more quickly (unless it is
not 'good enough' from an accuracy perspective)?
I agree in an ideal world, ground-surveys done from first principles will
be preferable. But the fact is that there remain areas of the country that
routing. Wolverhampton and Newcastle for instance are places which we
certainly would love to see become usable for routing as early as possible.
I agree with other posters that mass, automated imports sound like a recipe
for disaster, from both a technical and, crucially, a
volunteer-encouragement perspective. So wouldn't the pragmatic way forward
be to patch in / trace over data in incomplete areas, so that it forms the
basic network? Surrounding map features can be augmented on top of that
more easily on an incremental basis, leading to having the same richness as
other areas (e.g. Cambridge, which is an amazing level of detail, and which
forms for us an excellent testing ground for our engine because we know the
data side is pretty faultless).
Might it not be a best use of volunteer time to be adding really
interesting map features that other sources do not have, rather than having
to use this time to get basic streets in place first?
We certainly would love to see (more universally) stuff that other datasets
don't have, like whether a place is lit or not, but it's going to take a
while to get to that if basic street patterns still remain to be
walked/cycled first.
> If people want other sources so much, won’t they encourage the effort to
> combine the independent sources?
I think that surely underestimates the effort required. Processing one
massive dataset is bad enough, never mind having to do this for several
sources and find ways to deal with merging/clashing.
http://www.cyclestreets.net/journey/help/osmconversion/sieve/
shows our current sieving routine (NB Andy's pointed out that there is a
better way to do this upstream, but we've not had the time to migrate yet.)
OSM is a great project, and I think the head-start of basic road data,
which can be adjusted and improved on, is something that is needed to get
towards completion more quickly, creating a superset that will become
irresistable to people that were previously sceptical about open
geographical data.
Martin, ** CycleStreets - For Cyclists, By Cyclists
Developer, CycleStreets ** http://www.cyclestreets.net/
More information about the Talk-GB
mailing list