How to consolidate different API needs for different backends?

[pllim]

One of the things we had discussed before but didn't resolve was how does this API consolidate backend differences? For example, glue needs extra things like "reference image" and so on for linking, which no one else cares about.

A more concrete problem is like this: If Imviz has 2 images linked by WCS, but you use the markers API and input only has pixel coordinates, you need to tell Imviz which image the coordinates are for. This is further complicated by the fake that Imviz does create a hidden "fake WCS" image that is only 10x10 when WCS linking happens, and makes it the "reference image", which is useless for all pixel inputs, so you have to provide a non-reference image for the pixel coordinates. I hope this makes sense.

There are other problems that would require special inputs for Imviz here and there but I cannot recall right now. The gist is, whenever one of these situations arise in any of the backend that adopts this API, how are they supposed to implement this API while still make things work for their backends?

[mwcraig]

I think that this is addressed by #45 and #46, which between them add a way to specify which image an operation is on and a analogous way to track catalog operations.

[mwcraig]

@pllim -- closing this because i think it has been addressed (except for allowing kwargs everywhere). Please fell free to reopen if appropriate.