On my digital camera, street lights that use Mercury Vapor light bulbs come out green on camera. For example....
General Electric M400R3
Crouse-Hinds OVM
General Electric M400R2
Crouse-Hinds L-150
McGraw Edison Unidor 175
Cooper OVX
or most recently General Electric M1000

To your eyes, it's fluorescent white-ish with a hint of light pink. But High Pressure Sodium, Metal Halide and LED come out correctly on camera. But on my digital camcorder the Mercury Vapor luminaires come out right on camera.
Watch this video: http://www.youtube.com/watch?v=5vxuKWKZDyo

I don't get it.

The camera's spectral sensitivity is different than our's

Our sensitivity is mostly a wide area centered around green
In the merc we see : green and blue mercury peaks + red phosphor wide area
In the MH we see : many close peaks of the mercury and other metals

The camera uses matrix of sensors with individual sensitivity areas : red, green, blue

I guess that in the merc the green of the mercury falls exactly in the maximum sensitivity of the green pixels of the camera, so they are "high", while the other colors of the merc dont match well the red and blue sensitivity areas of the camera, so the camera sees them "medium" only. Result : med red + high green + med blue = greenish

In the MH there are way more peaks and way more complete spectrum, so there are peaks of red and blue (which dont exist in the merc) that match the red and blue sensitivity areas of the camera well, so all 3 colors are "high". Result : high red + high green + high blue = white

Different cameras can either have different sensitivity areas (as result of different sensors used), or differently working white balance (when the image colors are edited by the camera itself, in an attempt to get whiter image than the camera really sees)

It even seems to me, then some camera have build in correction for this effect.

Anyway it seems to me, then photographic industry didn't notice yet, then the light sources around are not only the black body radiators...

i noticed many cameras auto white balance is poor.
had a funny one after i rebuilt the entry light in a motorhome with nichia jupiter led's.
in person white with a slight violet tint.
took a pic at night with a samsung omnia and got a blue light and the campfire flames were green!
got a shot direct on to capture the prismatic lense effect where 3 led's look like 12.
then it came out right.like the wb gets confused easily.

The issue with white balance is, then it is capable to balance out only the color temperature (so the red vs blue, but not the green), so it work well only with hotter/cooler incandescent sources. And it is more then 80 years, when people don't use incandescents only, but still nobody program that correction into the cameras...

You're talking about AUTO white balance, most newer digital point and shoot cameras have preset settings for different light sources and some even have a manual preset mode where it sets whatever light source you point it at as white. And if you need more choices well get a professional level camera.......

But even that "point at white" balance very frequently is able to compensate only the "color temperature" and not the other color shifts. I do not understand that, because the generic correction for any shift would be way easier to implement (simply find coefficients for all three sensors, so you get equivalent output after the correction) then following the color temperature only (you have to implement quite complex function with single parameter).

In this matter cheapee cameras perform ironically better, maybe they programmed them so to correct mainly for the imbalance of the sensor alone. But more expensive cameras usually fail in the ability to balance a scene lit by MV or even some LED's...

well on a cellphone camera i should not be too surprised.
my droid 2 gets colored banding on live fluoro shots.i now have a canon rebel xti so i should be able to do much better now.

But even that "point at white" balance very frequently is able to compensate only the "color temperature" and not the other color shifts. I do not understand that, because the generic correction for any shift would be way easier to implement (simply find coefficients for all three sensors, so you get equivalent output after the correction) then following the color temperature only (you have to implement quite complex function with single parameter).

In this matter cheapee cameras perform ironically better, maybe they programmed them so to correct mainly for the imbalance of the sensor alone. But more expensive cameras usually fail in the ability to balance a scene lit by MV or even some LED's...

Not necessarily I've had good results with this method, sometimes I think you drift more towards the theoretically instead of whats more practical for the average user....