When placing new tracks that need to join with existing tracks I have noticed some differences between the existing OSM feature and where I believe it should be (deliberately avoided using the term ‘error’). In some cases this makes it impossible to join new tracks with existing tracks.
The tracks that I have seen have been very coarsely mapped with large spaces between waypoints so I assume that it is OK to ‘refine’ the work of others to make the map fit what I believe to be correct to match GPS data. Most of my GPS data has two traces recorded when going in opposite directions - so I am not relying on a single GPS trace which may have inaccuracies due to signal reflection or absorption.
I don’t want to destroy the work of others simply because I am the last person to edit the map, so I was wondering how others approach this problem and what is the correct OSM etiquette for this situation.
Only edit our own tracks and where it is impossible to join new and old tracks …
a) leave the tracks disconnected
b) connect them so that they can auto route, even if the track shape is different to the one on the ground
‘Correct’ existing tracks as long as we have GPS traces to back up our claim that they are in ‘error’.
Contact the person that entered the track that is in conflict so that they can edit or give permission to edit their work. (How do we find out who last edited a point and how to contact them)
Report the discrepancy to some higher source to fix the problem.
This is always a difficult issue. You are right, one ought ot be very cautious with modifying work from others.
However, there are reasons to do this:
new and better data have emerged. Old data (say several years ago) may have a very low resolution.
original data were gathered with points very far apart and have been “interpolated”.
and others may be there as well.
What I usually do is keep track of the accuracy of my GPS while gathering data. If the data differ less then twice the error, I leave as is (this is usually on between 10 and 20 m for GPS tracks with EGNOS/WAAS enabled). If it is more, I make a modification of the original data. Else I give the original data the benefit of the doubt.
Also I download other GPS data from the database and see how they compare to the data in the database.
I’d only correct the original data if I had sufficient information to justify a change. I wouldn’t assume my GPS trace is any better than the original source. I’d only change the original if there was enough GPS traces to justify the change (from different users) or there was something obviously wrong with the original data - for example when you stand in the road you can see it is straight but the the data suggests a zig zagging road for example.
As to problems connecting your traces to the rest of the data. I’d probably favour connecting it, even if the connection wasn’t perfect. I tend to view the connection of more value than having a road in slightly the wrong place. In this case if the necessary correction was more than minor then I’d probably add a FIXME tag with an explanation so that somebody can see the place needs investigating.
All of the above I too use to decide if I change somebody elses work, but I use another method on top of that: If I see an error, I first check the history of the element, if it’s been touched within a month I tend to leave the error be, thinking the contributor might still be active in that area and will notice the error himself and correct it. If it’s a few months old, I let it depend on the severity of the error. If it’s close to a year or more, I don’t hesitate to change it.
Sure, connecting new roads to existing ones is necessary. From experience, I usually assume my GPS trace is reliable. And I verify that by comparing my trace when moving on existing roads in the vicinity. If the same trace is fine in the neighbourhood, but is much different to the exisiting roads in an area where I want to map new streets, then what can be the conclusion? And of course if you see a straight road being zig-zag, this is strongly indicating that the existing trace is not accurate, or better say it’s less accurate than your own one. So: Minor differences you can leave, big and obvious inaccuracies I would always correct.
And, please: Don’t see corrections and additonal features added as “destroying other’s work”. Unless you completely delete all the previous streets just to paint your ones there, you don’t destroy anything.
Downloading exisiting GPS tracks of the area is a good idea, too. Then you have an average track. Always keep in mind that some roads could even be created by “repainting” satellite or aerial photos - these are often inaccurate or extremely wrong positioned.
I feel a bit rude now, as I rarely feel any concern at editing things other users have done. I usually spend a few minutes thinking about why they have done it the way they have and sometimes I learn new things that way. If I disagree with what they have done I’ll change it.
Sometimes I store the previous state in the note tag.
That’s almost unnecessary now we’re on API 0.6 but I still do it anyway.