[Openicc] Printing Plans GhostScript / sRGB / ICC

Chris Murphy lists at colorremedies.com
Thu Mar 3 11:18:34 PST 2011



On Mar 3, 2011, at 3:12 AM, Richard Hughes wrote:
> In which case, the devices are *not* going to be sending
> device-specific data at all. They'll hopefully be tagged PDFs where
> the images are tagged with common spaces like AdobeRGB and the text is
> assumed to be sRGB. In this case, the printer device that is accepting
> random PDFs over WiFi or whatever will hopefully have a half decent
> PPD file with a linked ICC profile in cupsICCProfile all present in
> the device ROM.



> In that way, we don't need embedded device profiles in the PDF, or to
> install colord or oyranos. The only time we care about PDFs with
> embedded output profiles, overrides and user calibrated devices is on
> the "full fat" devices like PCs. Of course, I'm sure for the next
> decade all mobile devices will send documents without any color
> management, and we just have to assume they are sRGB, which for the
> typical use case of most embedded devices[1] is probably fine.

A big problem with the PDF spec *and* how applications create it, is /DeviceRGB is what's used by default and we don't actually have a /DefaultRGB thats used by default. So really, /DeviceRGB is polluted to the point that it's ambiguous. We can't expect myriad mobile applications all over the world properly create PDFs for printing, properly tagging the PDF, unless there is an API that writes out the PDF, not the app.

On Mac OS there is such an API. No application writes PDF at all. This is done by the QuartzPDF Context. And it ensures there's no such thing as /DeviceRGB (except in a certain situation that I consider a bug so I won't get into it). If the app doesn't specify a color space for each object, then this API writes out sRGB ICCBased for all objects in the resulting PDF.

What would be nice is if web browsers did a better job of honoring EXIF and passing that metadata through to the print stream to the system, so the resulting objects in PDF were properly tagged, and hence displayed and printed.

Chris


More information about the openicc mailing list