[OSM-talk] The Highway tags and other junk strikes back

[Andy Robinson]

This thread seems to be getting a little messy so I thought I'd drop in and
give you my thoughts.

With respect to the way in which features are created in the data set we
must remember that OSM uses a very simple but effective model for the format
of the data. 
I look at the differences between traditionally drawn maps (cartography as
an art form), GIS and Engineering Drawings. Engineering drawings detail the
precise nature of every feature they portray as lines. So, for example a
gate post perhaps is drawn in plan view as a square of four lines (which of
course can also be considered as a box) precisely where the boundary lines
of the post exist for the slice elevation drawn. In a traditionally drawn
map, and if the scale were right, then the cartographer might draw the
gatepost as a square but equally it might be represented differently because
the cartographers art is not necessarily to portray the features exactly as
they exist.
GIS sits somewhere between the two because it takes information that is
created from modern CAD formats (where they are generally called "edges" in
GIS) and digitised features produced from analysis other data sources such
as traditional maps.
So how does OSM fit in here?
My view is that OSM draws from both input routes but does not try to be a
traditional GIS structure. We could in theory add all our road edges and
gate post boundaries as edge lines and render them accordingly, but that's
taking the data entry beyond what OSM was envisaged for. OSM's basic
assumption is that a linear feature is represented as just a single line and
that this approach is extended to logically discreet features that fall
naturally as nodes and those which naturally encompass something and
therefore become areas. The current one major exception to the single linear
representation is perhaps for the wider rivers were establishing a common
width for a section is not really practical and where the use of a riverbank
feature tagging has been found to be more useful.
Therefore my belief is that OSM should not try to shoehorn itself into the
traditional GIS approach but rather should always be considered as GISLite.
With appropriate tagging its possible to achieve with the OSM format
something approaching that which can be achieved with a full GIS model, but
it will always me missing some of the more sophisticated tricks.

With respect to Tags, I've commented many times before about my views
regarding the extensibility, that's why I wrote at the top of the original
Map Features page that anything goes for tags and values as anything else
stifles what the project must maintain...creativity. Many of us benefit from
"standardised" key/value pairs as it enables these to be seen by rendering
software and other third party applications and used in a common way, but
that's only part of what OSM is all about. Therefore I hope everyone will
continue to respect the tagging of others. If someone tags something with a
key value pair that is not listed on the wiki, or anywhere else, it should
be equally valid use to that of a key/value pair that's been accepted as a
"standard" via proposal and voting. OSM is not about restricting the volume
of data it holds, on the contrary I believe its about the opposite, though
clearly as time goes on the tools required to retrieve the data you are
interested in rather than all of the data present in a bbox will need to be
improved. 

Cheers

Andy

Andy Robinson
Andy_J_Robinson at blueyonder.co.uk