Instead of the only options being 'obviously wrong' and '90% correct', how would you feel about adding some presets to use the matrix tags from your standard sRGB or Adobe1998 profile (or possibly a parameter to specify any ICC profile with matrix tags) in order to simulate a more a 'neutral look'? By 'neutral look' I mean a look similar to color management being disabled (if your working color-space and image profile matched the tags you attached). My logic being, this would give you a viewable image, unlike the 'obviously wrong' matrix tags without fooling yourself the image really is being color managed accuratly with the '90% correct' matrix tags. Or is my logic flawed in some way, and this won't work or is a really bad idea for one reason or another?
I really don't think it's a very good idea, because I think in many situations people won't notice anything immediately if it's not obviously wrong. It will be some way down the track if at all that they notice something strange is going on, and that will then spawn various queries and lengthy investigations to determine why the colors look sort of right but aren't actually right. I'm also not sure what you mean by '90% correct' - the matrix tags will be as correct as they can be made for a matrix profile, it's just that (by definition) they don't represent the same behavior as the cLUT tags. Generating a R<->G swapped matrix + XYZ cLUT and also a matrix + XYZ cLUT and trying each one means it's easy enough to figure out which one is being used. Not knowing what's going on is one of the curses of color management - most of the systems are made to be "easy to use", "no fuss", "just work" so you can't tell what's actually going on, or whether things are not working as expected and fallbacks are being used. Crude diagnostic aids such as deliberately wrong profiles are then needed to try and get a clue, rather than being able to know that it's being done right. Creating a profile with a passable matrix tags that aren't actually related to the device plus an actual device cLUT seems to be just asking for trouble.
______ That reminds me, Graeme, how would you feel about coding/releasing a standalone tool which can add/remove/import tags to/from ICC profiles as a nice complement to iccdump which can only read/export tags? From the sounds of it you may already have a tool(s) written which can do some of these things, since eariler you posted an ICC profile which you had removed the matrix tag from, and colprof is obviously able to write tags to an ICC profile. I don't believe any tool with those functions is publicly available anywhere, so there is definitely a gap to fill. Even if the add/import functions of such a tool only worked (i.e.created a valid output profile) with Argyll created profiles, it would still be a valuable tool to have available.
I got the impression that Kai-Uwe had created such a tool (ICC Examin), but maybe it can't edit. I'm afraid I used another much older tool that I have lying around (PfSnoop.exe from Kodak) to delete the tags, or sometimes I modify icc/iccrw.c to do a specific task. I've certainly thought about creating such a tool, but it is rather tedious, and I think I have far too much on my plate to contemplate this in the near future. If you'd like such a free tool, perhaps the best approach would be to contribute code to ICC Examin ? cheers, Graeme Gill.