Let me plug OpenLayers, which uses OSM and is way harder to use than google maps but once you stop hating on it you realize it can do a bunch of things that google maps really can't and is more versatile in numerous ways - ymmv as always. (https://openlayers.org/en/latest/examples/).

I've found it more useful for specific applications that I build because it's far more customizable. The map truly feels like it's part of the app and not just some mapbox or leaflet thing sitting on top of gmaps. Again, these aren't hard facts, just stylistic preferences on my part.

The ES6 is not really a dealbreaker if you're using something more traditional. I just make my own "map.js" that exposes the interfaces I need, then I transpile it and include it like anything else.

I know that sounded really complicated, but my package.json is 10 lines, that's it.

This way you can use it in a much more conventional classical way without jumping the whole project over to ES6 syntax.

I find that if you need more complex mapping features then OpenLayers is really good. It can do an amazing amount straight out of the box, but it does take time to get used to the API. It is more like an actual mapping or GIS program.

But for building functionality outside of the map I tend to use Leaflet instead due to the ease of use. You can treat the map more like a black box and the API is easier to use.

Surprised that no one is talking about ways to leverage the server so you’re not doing all the work on the client. You can shift layers by precomputing either raster or vector so you don’t push your poor, maybe mobile client to the max.

Classically you would do this with Tilemill for rasters which was amazing software, if crashy. And since we’re on a mapbox thread what I think put them on the map. I wish it was still actively developed and adapted for vectors.