CMYK exists because, though is possible theoretically, it isn't possible to generate black from mixing CMY inks. As the C, M and Y inks aren't perfect and have some contaminants, mixing them ends up in a dirty brown instead of pure black.
That's why CMYK exists, and that's why it isn't so simple to print an RGB image.
The problem resides mostly in the generation of black and gray shades. Although there are systems that do a great job converting a photo to CMYK on the print side, it's still a problem to do a simple task as placing a pure black overprint on a solid color background without destroying some underlaying information on the separation. I'm a designer, not a photographer, and an image manipulation program is an essential part of my workflow. And placing some black text, or editing a large image with a black or gray background, adding black drop shadows, aren't rare at all. And it's a pain without the ability of editing the separated CMYK.
It's not about if the printer will handle the file or not, is about creative control. Sometimes you NEED to control the black overprint. Sometimes you need to use spot colors and you need to control the channels and how they overprint.
Even with Adobe software, before having spot channels in Photoshop, it was a common practice to separate to CMYK to make 2, 3 or for 2 inks prints (replacing the corresponding plate's ink for a custom ink). Simply because other programs didn't support the Adobe's custom multitone files but did support CMYK tiffs.

Well, I can't do duotones with Gimp to insert in a 2 inks flyer. CMYK editing would help. I can't control the black generation of a separation, because the separate+ plugin doesn't support that. It just support existing profiles and there is no control. I can't control CMYK curves. And trust me, that's extremely common.

I can live without CMYK, I have some workarounds and can do some tricks, but it certainly makes my designer work harder and less productive. I can wait, I'm ok with the argument "it's not trivial, requires a lot of work, requires a lot of refactoring". But I'm not ok when somebody says that it isn't necessary.
Maybe it isn't for photographers, but it certainly is for designers. GIMP stands for GNU Image Manipulation Program. Not just for Photograpy. I think that CMYK editing is certainly in the scope of the product vision.

Hi Robert, thanks for your comments.

This really sounds like you're using black as a spot color rather than going generic editing in CMYK space.

That was just an example. Another example could be just putting an image in front of a gray gradient background. In my experience, it's not that easy to find a printer that can convert a RGB gray into a perfectly CMYK neutral grey. Or isolating a photograph, putting it on a solid color layer and apply a drop shadow of the isolated image on the color. Why should I send an RGB file and see how my printer's RIP separates the grays and the shadows when I could be specifying: I just want 100% black over the color?
They are only real world examples. Probably there are excellent print shops in Germany or USA that deliver excellent results from an RGB jpeg, but that's not always the case.

I question whether doing this in an image editing application is really the right thing to do. If you're doing black text, you probably want the text to be vector rather than raster anyway -- printing an image scaled to 240 DPI is fine, but the text won't look so good at that resolution. In that case, you're better off using something like Scribus to do that kind of overlay, at least until GIMP has vector layers.

Again, that was just an example. It may be true in a brochure or a magazine page, but what if you need to add a texture to a title, breaking the borders of the characters with a "grunge" brush? or something like that?

But let me show you a very simple example: http://www.flickr.com/photos/superdd/2781429410/ This image was created in Inkscape and exported as PNG. Then in Gimp the yellow part got some texture and color work.

What if I want to put that image in a page of a magazine that I'm creating in Scribus, and I want the black part of the image to be the same black (100% black) than the text. In that case I would have, according to your workflow, to: -take the image to gimp, make the texture part, remove the black part, save it, import in scribus the color blob and a black-only SVG version of the drawing, make them match in size and alignment (ouch, I should have imagined that I would need some bleed on the color shape to avoid alignment errors), and export them as a CMYK PDF to send to the print shop

Or simply separate in GIMP to CMYK, remove the black part of the CMY channels and tweak the black channel, save as TIFF and import in Scribus.

Call me crazy but I choose the last one.

Of course there are alternative ways, but sometimes it's faster and more direct, thus preferable, to do it with the image manipulation program that using three or four separated applications for a simple task.

Which again is a spot color kind of use case rather than a CMYK use case.

Yes, but we don't have spot channels either. At least having CMYK would work as a viable temporal solution until we have spot channels. I find it hard to imagine that GIMP will support spot channels if it won't support CMYK channels. It wouldn't make much sense to add a spot channel to an RGB image, would it?

Does that indicate that separate+ is what needs to be enhanced, rather than the core application?

This discussion was about a PDF export plugin at the beginning. I was trying to make evident that a PDF export plugin is probably useless without CMYK/SPOT/VECTOR LAYERS, and improving and integrating the Separate+ Plugin instead of focusing on a PDF exporter would make sense and a big difference.

Gez