A little more research showed TTL metering works with a linear, but not properly. The problem is that if the polarized rays hit the semi-transparent mirror at the wrong angle, the light will not be diverted properly to the sensor resulting in strange, even blacked out, readings. You can still work around this by using MF and bracketing exposures (start out at a one stop increase from the non-filtered exposure), if you don't really want to go out and purchase a new filter at the moment. The photographic results should be the same it's just more hassle. (it turns out that because linears do not have an extra wave plate the image quality is generally better)
I used to think exactly as you do about the differences between linear and circular. But it turns out that they polarize light exactly the same way (like blinds). The difference of a circular filter is that after the polarized light is filtered it is passed through 45 degree quarter-wave plate that separates the waveforms. Instead of waves oscillating the same direction, they are sent oscillating in a circular pattern.
This stuff is really interesting. The designers of polarizers sure know what they're doing.