3 ms·
When you're down to 90% of them only providing location (unless you stop caring about topology), it's no longer as clear-cut. I was only really arguing against
by ghusbands 2y ago
When you're down to 90% of them only providing location (unless you stop caring about topology), it's no longer as clear-cut. I was only really arguing against the implication that it's "funny" to use nodes; the topological connections and normalisation/deduplication do have advantages and it looks like a reasonable technical decision to have made.
I also imagine that creating new topological connections may become trickier when most locations are 'hidden' in ways.
However, I do hope you find a common way forward that reduces the apparent redundancy and excess processing/storage requirements.
- zigzag312 2y agoI found the added complexity of this normalization funny/ironic when I wanted to process some OSM data. As IMO, the benefits don't nearly outweigh the added complexity and costs caused by it. A different way of creating new topological connections wouldn't be that difficult. For example, some topological connections are already being created via relation type. I'm not involved with OSM project, but I think that some changes to OSM data model would benefit it greatly. Resolved geometries would enable stream processing and decrease complexity for majority of task. It would make OSM data much more accessible.