Re: [overpass] error with attic data

[mmd <>]

> 

> - preprocess history to wrinkle out such inconsistencies. But then my
> instance will lie about some objects history.
> 

Did you already run some analysis, how many objects would be affected by
these inconsistencies in Switzerland? Also, what would be the largest
time difference needed to be applied to an object to make it consistent
again? In your example, that was only a matter of two seconds.

So my questions would be:

  a) we're only talking about few objects
  b) the time difference is just a matter of seconds/minutes

I wouldn't bother tweaking the overpass update logic and fix in the
inconsistent input data by some automatic pre-processing instead.

This seems like an acceptable approach to me, as your use case looks at
longer term OSM contribution development anyway.

Maybe Jochen's osmium library offers something in that area already.

-- 
cheers,
mmd