[Tagging] RFC - Feature Proposal - area of steps for pedestrians.

[Christoph Hormann]

On Saturday 04 May 2019, Tobias Knerr wrote:
>
> Here's the raw data if you'd like to examine it:
> http://tobias-knerr.de/upload/Step%20Polygon%203D%20Examples/
> Please excuse the sloppy mapping, those are just intended as tests.

Thanks.  I guess that means your approach relies on a one-to-one 
relationship between linear ways and polygons - or at least on the 
polygon outline being intersected by the linear ways exactly twice.

> While you correctly point out several further limitations, I think
> it's important to keep in mind that this isn't an attempt to define a
> data model that works for everything. It's about finding a sweet spot
> that works for a sufficiently large class of steps to be worth it and
> is still relatively simple.

What would likely happen with your approach is once there are data users 
using it mappers would start splitting any stairs that are not 
supported by the algorithm artificially into parts small and simple 
enough for the algorithm to deal with and you'd end up with data 
modeled for the specific requirements and limitations of the algorithm 
rather than for mapping-efficient documentation of the nature of the 
stairs.

> As for that data model that works for (almost) everything, I believe
> that will have to be drawing a way across the edge of each step.
> [...]

That would essentially be giving up on mapping the stairs as an overall 
feature and modeling the steps individually instead.  For data users 
that need only individual steps in the end anyway this would be 
convenient but for any data users who want to interpret the steps as a 
whole this is a bad idea - likewise for the mapper who does not want to 
mechanically draw step after step.

Circular concentric steps for example like this:

https://www.designdiffusion.com/en/2018/10/15/japan-plaza-cofufun-designed-by-nendo-for-the-local-community/

you could model with:

* one circular way per step.
* a number of sector polygons and centerline ways suitable for your 
algorithm.
* a single node and tags for step count, height, inner radius and outer 
radius.

-- 
Christoph Hormann
http://www.imagico.de/