[OSM-talk] Source tags and changesets (was ... ODbL ...)

SomeoneElse lists at mail.atownsend.org.uk
Fri Jun 17 11:57:07 BST 2011


On 17/06/2011 09:42, Jochen Topf wrote:
> The source tag can sometimes be some help in figuring out the history 
> of an
> object. After it was entered in OSM we have a complete history, before that the
> source tag can sometimes help. But it is far less useful in practice than many
> think. Exactly because the data almost always is a mix of many sources. In many
> cases source tags on objects are even misleading, because people don't change
> them when they change data, so the data only shows one source when it has many.
> Its better to not have them at all.

I'd agree that people sometimes don't update source tags on objects; 
just recently a non-local armchair mapper updated a bunch of stuff 
locally and did exactly that (they apparently put a source tag on a 
changeset, but as that's not visible against the object that's not 
actually useful).

I'd disagree that "it's better to not have them at all" though - the 
more information about what's there and who added it the better.  When I 
create Garmin maps I currently do the following:

o Ways and nodes from old out of copyright maps are labelled as such, so 
that e.g. footpath data inferred from there can be updated with actual 
rights of way

o Ways and nodes added with a remote source or by prolific tracers from 
OSSV and Bing are labelled as such, because any "on the ground" stuff 
will have been missed there

o Ways and nodes added by, er, "users with historical accuracy problems" 
are marked as such so that they can be explicitly checked.

So yes, sources get combined (and you'll see lots of semicolons in 
source tags from me reflecting that) but personnally I find source tags 
on OSM items extremely useful.

Cheers,
Andy




More information about the talk mailing list