That may work as well as "ideally, applications should be agnostic to language and an automatic language-transformation layer should handle all the text".

While it's technically possible (several app stores do just that), the results are far from good quality. When colors are used to convey meaning, the transformations needed to allow blind-color users to see them will be application-dependent.

This may be enough to allow these users to play simple games, but ideally developers should learn enough about the human API so as to cater to the real needs of their users. Depending on invisible colors for gameplay should be seen as bad as letting an uncaught exception crash your application.

At least with Chrome you can install the Daltonize! plugin. That plugin even has a "simulation" mode to help color-visible to see how your page looks like to (some of) the colorblind out there.

AFAIK, colour blindness ranges from barely an issue to an almost complete "grey-scale" style of insensitivity to colour. In general, the only safe design choice is ensuring that no decision-critical information is conveyed through colour alone.

Only if you want to catch 100% of the problems I'm slightly colorblind and have had a hard time on the same level as the parent poster. Greyscale colorblindness (i.e. absence of color receptors) is extremely rare. PopCap does a colorblind safe mode (for zuma, peggle, etc), so did EA for Battlefield 3. It might not be the best solution, but it's good enough for 80% of the colorblind. There is a downside, a midway solution, although more realistic, reduces attention on the bigger problem, the rest of the colorblind population would be less catered to, but imho that already is the status quo and will be for the foreseeable future

So, service idea: pass a complex color-blindness test (on your setup) and it will generate a color profile suited for you (and your setup).

I wonder if it is possible without ruining the appearance of everything else (never really looked into how color profiles work).

The bad thing about this idea is that it wouldn't be reliable for the user's eye because of the variation of color representation across screens.

The good side is that it would be a correct representation of the accuracy of color recognition of the system screen-eye, which is much more relevant from a design perspective.

I have always wondered about this.