[OSM-talk] An open letter to OSMF board members.

[Tomas Straupis]

2022-06-04, št, 06:55 Paul Johnson rašė:
> The Standard map seems to be very intentionally a generic, culturally neutral
> <...>

  Everything in this world must have some purpose, some target (could
be a target in Laozi sense). When OSM data is so diverse in all
dimensions, you simply cannot have one DEFINITION of what "generic
map" is, therefore it is impossible to create a "generic" map which
would satisfy most users. When one creates a map, one has to answer
some questions: what/who am I creating this map for? who are my
readers ("everybody" is not a good answer)? what [information] do I
want to communicate to them? You do not simply drop a pile of data,
put some style sheet over it and say "here it is - a map!".

  Questions for thinking:
  * can we have playgrounds and casinos on the same map?
  * can we have bicycle pois on a map and yet have motor transport
ways overlap cycleways (graphically or by visual importance)?
  * what should be more visible for cayaking map - a river or a road?
  * Or even getting to some hardcore: you have a hierarchy of
settlements on your medium scale map (not all settlements are
displayed because of lack of space), would you like to fine-tune the
hierarchy and display even small settlements even on smaller scales
because those settlements are important for your specific map?

  So in general: visual hierarchy, importance of the same classes is
different depending on the map you're creating.

>  But please take the nearest style to what you're looking for, fork it, and
> develop it closer to your vision.

  Paul, this is what I've been doing for some ~10 years. We have maps
for tourism, cycling, craftbeer, topo, cayaking as well as some
technical styles for showing what a road or administration hierarchy
is (and I do not mention paper maps which are a different story).
Students are playing around with this data creating even more maps out
of it. Therefore, what I'm writing in this thread is according to my
experience with creating those maps.

  Not only there is no single "generic" map, even more - there is no
single generic vector tile source/dataset (unless you want to make
some very impractical size tiles). You need to divide the data you
want to provide (in vector tiles) into logical groups finding a
balance between too many different tiles being loaded for one map
(inefficient) and huge tiles with a lot of unused data (also
inefficient). There are a number of technical solutions for that, some
are already developed, some are not (but this is another topic).

  One lesson is that wile you can get away without generalisation for
raster tiles, you will have to generalise for vector tiles (or will
have to remove some important layers on pretty large scales).
  Question for thinking: how much data does a road with a length of
one pixel takes on a raster tile? how much data does it take on a
vector tile? what if you have hundreds of overlapping one pixel roads?
(here "one pixe" - a road which length is one pixel on some specific
scale and dpi).
  But cartographic generalisation was always an unwanted step in OSM
for one reason or the other...

  Getting back from vector tile topic to the topic of the thread -
having one raster style is a dead-end. We either have that one raster
style clearly defined as "not for general use", or move to vector tile
sets (multiple? different entities providing different tile data
sets?).

-- 
Tomas