Joe Amenta

AFAICT, this is a seriously complicated problem. I've pasted a completely untested *START* of how I would go to deal with holes outside shells at the bottom of this, which...

> Seems like a union of the collection items should be done and the comparison done on that? I don't think so. Even if the elements were just `Polygon`s, consider...

> @airbreather is there an upstream issue or branch that we can follow the progress on? @archanox opened locationtech/jts#539

> I think we shouldn't lock up our library so no external extensions regarding geometry types is possible. `InternalsVisibleTo` is only necessary for `internal protected abstract int CompareToSameClass(object[, IComparer])`. I...

> All the issues you listed above can be solved/approximated by calling `ICurve.Flatten([arcSegmentLength])` prior to invoking them That doesn't solve the problem of NTS core needing to have built-in support...

Reported as locationtech/jts#497

I think I'd like to keep it open unless we decide for sure that we're not going to do anything about it until JTS does. `STRtree.Remove` and `Quadtree.Remove` look like...

As an aside, the recommendation of multi-targeting .NET Core 3.1 is understandable, but unnecessary: a member of the community has stepped up and created a NuGet package called "[Nullable](https://www.nuget.org/packages/Nullable/)" that...

Note: using https://github.com/tunnelvisionlabs/ReferenceAssemblyAnnotator instead of the "Nullable" package, since the former does the same and also gives us straightforward access to the nullability annotations that the .NET Core team has...

Can you please provide the code and data that you're using? Without those, I can't prove that there's even a problem anywhere near this code, much less prove that changes...