I’ve doing some analysis on historical data (specifically Geofabrik’s New York State full-history extract using Osmium Tool) and found something somewhat surprising.
https://www.openstreetmap.org/node/42428272/history
The first version of the Way with ID 5669733 was created on 5 September 2007 and contains the Node with ID 42428272. The first version of the Node with ID 42428272 was created on 28 September 2007. If I understand correctly, the Way included a Node which did not exist yet. If this is indeed the case, how does this sort of thing work? I’d like to build visualizations of New York overtime, so understanding the historical data to a tee is important for my purposes. Thank you.
1 Like
If you go back that far there are all sorts of anomalies in the OSM history data, because a) there were bugs in the code so that wrong data was occasionally created, but more importantly, b) the data model of OSM changed over time and you might be seeing some artefact of that. In September 2007 API 0.4 was used, which still had segments, so the data model looked quite different. The data you can download now was converted from that old data model into the new data model at some point and that process might have introduced anomalies.
If you want to understand more you’ll have to read about old API versions and find the original old data. But generally it is not worth it to look at any data before October 2007 when API 0.5 was introduced, because there isn’t that much data to begin with before 2007 and API 0.5 was quite a big break compared to API 0.4, later changes were not that big, so the data compatibility is better then.
If you do go back and find out all the details and anomalies, please share them with us. You are not the first person to try this and being frustrated by things not adding up. 
2 Likes
Ah, I understand. After doing some more analysis it became apparent that a truly minimal amount of data has this sort of anomaly anyways, so I’m just going to move forward discarding anything suspicious. I’m glad I was not getting something wrong. Thank you very much for the quick reply and I will be sure to share any further findings.