The concept in bug 691607 can be similarly implemented when using the clist with devices designed to optimize rendering to halftoned (CMYK) devices such as the wtsimdi (or a later, better implementations). We know that often there are large areas of the page that can be painted without needing non-idempotent ROP effects. If the ROP is idempotent, then the object could be painted using the methods to paint halftoned data directly to the non-contone buffer (e.g. CMYK). This will also reduce the amount of memory needed for the contone buffer, and reduce the area that we then have to perform color conversion on in the raster domain (RBG contone -> CMYK contone, then halftone). The smaller the area we have to do this with, the faster a non-ASIC color PCL printer can be. The problem is that ordinarily we don't know what area can be painted directly to halftoned CMYK an what area must be painted into the RGB (or RGBtag) contone buffers (other than the band by band 'complexity' data we now collect during clist writing. When we are using the clist devices, however, we aren't actually doing any painting -- operations are simply going into the clist. If this device collects a 'bbox' of the area actually needed to be handled for non-idempotent ROPs during playback, then we can use this information during clist_reading to allow the rendering devices to constrain the bbox of the contone buffer. The clist playback would then render objects using the halftone buffer if the object is outside the ROP bbox, or only to the contone buffer if it is entirely in the ROP bbox, or to _BOTH_ if it overlaps. It won't hurt if the objects that cover both areas are painted into target device's buffer since these 'marks' will be painted over when the contone buffer is color converted and written to the halftone buffer needed by the device. For pages that only have a small area affected by non-idempotent ROPs, this can speed things up, and may reduce the memory because the full width won't be needed for the contone buffer in bands that do have some difficult ROP operation. In order to implement this, we might want to make a ROP_compositor device that more closely mirrors the pdf14 compositor device to 'normalize' the way we handle this selective playback from the clist.
Assigning this back to Ray. It is likely that he could do what he did with transparency to what is needed here.
After 11 years, I think it's time to admit we aren't going to do this.