For PFS cameras, the slit head is the only light source, so we only ever get lines up and down the detectors. In order to build flats, we dither the slit head side to side. My flat creation task therefore needs to know about the slit position (I coadd images taken with the same slit position in order to remove cosmic rays, normalise out the fiber profile, and then sum the normalised images from different slit positions). I figure I can create a dimension that will let me do that grouping, but I need to get the slit position into the registry. I propose changes to the RawIngestTask and ObservationInfo that will allow me to subclass and extend their behaviours in order to accomplish this.