[Imports] Uploading sidewalks in San Jose, California, US

Andy Townsend ajt1047 at gmail.com
Fri Oct 20 11:18:20 UTC 2017

On 20/10/2017 04:43, Nick Bolten wrote:
>  It will never be the case that the entire pedestrian network is added 
> all at once, or without disconnected portions.

I'd suggest that adding pedestrian routes without disconnected portions 
is entirely possible.  It needn't be the whole pedestrian infrastructure 
for a city or county, but just adding a bit at a time, and ensuring that 
each bit newly added it routable is entirely possible.

It's possible that during the process you'll find some disconnections 
and won't have the data immediately to hand and need to go out and 
survey, but SJ is relatively small, compact and walkable* so that really 
shouldn't be a problem.  An example of where data has been added and not 
immediately connected is at 
http://www.openstreetmap.org/#map=19/37.33469/-121.91613 - to me it 
wouldn't make sense to add sidewalks without joins, because you'll want 
to use routing as QA to make sure that everything links up properly, and 
you can't do that if things "aren't supposed to link up yet".

Finding and fixing "unjoined way" routing issues after the event is far 
more problematic than getting it right first time (I've had to do that 
with lots of footpath data south of Manchester in the UK a few years ago 
and figuring out "not routing because of access rights" and "not routing 
because of a join error" is tricky)

> On the router end, it should be as simple as either removing 
> disconnected subgraphs or preventing disconnected subgraphs from being 
> selected during the initial 'find the closest valid way' step.

It surely makes no sense to suggest that routers should change their 
code to work around some incorrectly added data to OSM ...

Footways have pretty much one job - to support pedestrian routing (for 
various values of "pedestrian", obviously).  If they don't do that they 
might as well not be there.

> I'm in contact with the mappers putting in the time to map pedestrian 
> ways in San Jose and they're putting together a dataset of street 
> crossings to start importing. I'd like to suggest that we support 
> their endeavor and time commitments by supporting the inclusion of 
> street crossings, rather than discouraging mapping via threats of 
> reverts on changesets. Everyone's at the table and working to improve 
> the data.

Indeed, and there has been considerable supportive advice given (on this 
list and elsewhere).  Unfortunately what seems to be happening is that 
this advice is sometimes being ignored.  We should be aiming for a 
"proper, complete map", in a case where a partial import actually makes 
it harder to "map it properly" in the future then a revert of an import 
makes complete sense - though the situation at 
http://www.openstreetmap.org/#map=19/37.33498/-121.91646 suggests that 
some post-import patching up (sometimes referred to as a "bucket and 
shovel job") may be less work than "rework and start again".

Perhaps you could assure us that the mappers currently imported unjoined 
ways will, with some urgency, make sure that they are actually connected 
and that it won't be reliant on the wider OSM community to fix 
everything up after the event?

Best Regards,


* even by European standards - as an occasional visitor to the northern 
areas and the downtown for many years I found it more so than e.g. Portland.

More information about the Imports mailing list