[OSM-dev] Is this worth considering?

[Sandor Seres]

To OSM vector map-makers: do we need the 1:1 vector geometry on the map
server?

To be short I need to jump over many, many fine details. Yet I believe that
for most of the related notions and processes we have approximately the same
understanding. Anyway, in the vector geometry data preparation one of the
essential processes is the scale-levels generation. This phase is about the
end of the preparation chain, just before the (global) tiling.

I am intentionally using the scale-level notion instead of zoom-level
because the criteria - when and the model - how to create the scale-levels
is considerably different compared to the similar raster map-making
processes. For when, I use the data amount criterion (for now 11 levels) and
for how, I use a vector smoothing and data reduction model (sometimes called
data-generalisation). In this way for arbitrary scale factor the data amount
to transmit from server to a client is between 2 - 8 KB per km**2 equivalent
(for about 50 area and line OSM object classes). And here is my point:

In the vector mapping systems the start scale-level Lev0, created with the
scale factor 0.001 (1:1000), just perfectly replaces the 1:1 geometry. This
means that rendering/displaying the source 1:1 geometry over the geometry
from Lev0 (or the contrary) you will not see any differences on a 120 dpi
display surface. Eventually, you may exceptionally see one pixel difference
using high scale values like 1/200 or 1/300. But what is important that is
the data amount. Lev0 contains for about 15% less nodes/points compared to
the 1:1 level for the mentioned 50 object classes. In other words, using
Lev0 in scales between 1:200 and 1:5000, the server could handle about 426.5
million points/vectors less (compared to a bit more than 2839.3 million
vectors in 1:1 geometry). Of course this has impact on the transmission and
client rendering efficiency as well.

Thanks for the comments, Sandor.

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20150605/4916564d/attachment.html>