Currently when making references, we provide the RA/dec or the target/section ID as the identifiers for what part of the sky the reference should cover. We do this with a simple point in the sky and grab what ever contains that point.
However, it may be better to pass a specific image, so then you'd have its WCS and corners and everything, and you could align all the overlapping images to that target image. This helps avoid a situation in which the "most recent" reference image is not really close to the field you want, but you still align all images to that.
I'm not sure how that would fit into the provenance model we have for coaddition (where you define to which image on the list you want to align) and I am not sure who will provide the target image in this scenario... but it is something to figure out as we may need this functionality at some point.
Currently when making references, we provide the RA/dec or the target/section ID as the identifiers for what part of the sky the reference should cover. We do this with a simple point in the sky and grab what ever contains that point.
However, it may be better to pass a specific image, so then you'd have its WCS and corners and everything, and you could align all the overlapping images to that target image. This helps avoid a situation in which the "most recent" reference image is not really close to the field you want, but you still align all images to that.
I'm not sure how that would fit into the provenance model we have for coaddition (where you define to which image on the list you want to align) and I am not sure who will provide the target image in this scenario... but it is something to figure out as we may need this functionality at some point.