Consider alternative construction of globe multi polygon

[ajfriend]

We need a multipolygon that covers the whole globe for #1103, but there are several potential candidates. Consider, for example, using the H3 icosahedron faces.

See: #1104 (comment)

[ulekarsiddhant0-boop]

Hi @ajfriend, I am interested in this issue. So , please assign this issue to me.

[ajfriend]

Hey @ulekarsiddhant0-boop, thanks for the interest! We'd be more than happy to discuss a different construction of the globe multipolygon.

Just please do note that since we have an existing implementation (the 8 triangles) that might be "good enough", there's no guarantee that we'll want to move to something else.

Some tradeoffs I'm considering:

• the current triangles construction is fairly simple, which is nice
• the triangles don't have anything to do with H3, which is not so nice
• something like the H3 icosahedron would be nice because it is related to the library, but I couldn't see a way of doing that without adding a lot more code

So please do share any ideas you have in this thread and we can discuss more (before you sink a lot of time into a PR which may or may not be accepted in the end).

[ulekarsiddhant0-boop]

Hi @ajfriend , sorry for late reply due to my exams are going on and thanks for the clarification and for laying out the tradeoffs so clearly.

I agree that the current 8-triangle construction is appealing due to its simplicity and that it may already be “good enough” for many use cases. I also understand the concern about introducing significantly more code just to align the globe multipolygon with H3 concepts.

My initial thought was to explore whether a construction inspired by the H3 base icosahedron could be expressed in a relatively minimal way (for example, reusing existing constants or geometry already present in the library), but at this point I do not yet see a clean approach that would avoid added complexity.

Before investing time in a PR, I will think more carefully about:

whether there is a simpler H3-adjacent representation that materially improves over the current triangles, and

whether such an approach would provide concrete benefits (correctness, maintainability, conceptual alignment) that justify a change.

I will share any concrete ideas here first for discussion, as suggested, before attempting any implementation.

Thanks again for being open to discussing this.