Quebec is imposing a carbon tab, and according to this article (HT, NCPA):

The amount of the carbon tax varies according to the amount of carbon dioxide each fuel produces. For gasoline, the tax is 0.8 cents a litre, the charge for diesel is 0.9 cents, for light heating oil 0.96 cents, heavy heating oil 1 cent a litre, coke used in steel making 1.3 cents a litre, coal $8 a tonne and propane 0.5 cents a litre.

I suspect that a tax like this one will work better that a cap-and-trade system because carbon output (and sequestration) comes from so many sources. But a brief review of principles suggests that this is still a second-best approach. Recall this tautology: E = E/G x G/M x M, where E is the emission level, M is miles driven, and G/M is gallons per mile. This statement identifies three ways that pollution might be abated: burn each gallon cleaner, burn fewer gallons per mile, or drive less.

The infamous US CAFE standards sets a maximum G/M, making per-mile driving cheaper and probably increasing M -- a classic Laffer Curve problem. An input tax, a la Quebec, provides an incentive to economize on both G/M and M but not to achieve abatement efficiently if reducing E/G is the best way to proceed.

Since the target in Quebec is CO2 emissions, this is probably not a big deal, but why settle for second best? Anyway, why let the focus on CO2 take attention away from other pollutants?
The technology seems to be in place to tax emissions directly rather than inputs. This article shows how work by Donald Stedman points the way:

A remote sensor for measuring the tailpipe exhaust of moving vehicles, FEAT was created in 1987 with a grant from the Colorado Office of Energy Conservation.
Built around his Fuel Efficiency Automobile Test (FEAT) remote-sensing technology, the Smart Sign gives drivers instant feedback about their cars’ performance.
Stedman designed and built the first model, FEAT 1000, with help from the physics department and the University of Denver Research Institute. Research Engineer Gary Bishop created the software.

(I include the last paragaph mainly to note that Gary Bishop is a Berry College graduate.) Stedman's approach would allow pollutant-specific tax rates to be applied with direct implications for E/G values.

Is this important? This article by Lynn Scarlett suggests that it is:

Automobiles in general now account for as much as half of total air-pollution emissions in many U.S. cities. But these aggregate figures obscure a critical detail. Only a fraction of cars -- around 10 percent -- account for more than half of all vehicle emissions of carbon monoxide and hydrocarbons. It is these so-called gross polluters that California and other states are trying to target with their new smog check programs. ...
Targeting gross polluters sounds like a simple and sensible idea. The challenge lies in implementing it. Gross polluters are not just all those smoking old junkers easily identified by make, model, and year. ... Moreover, new cars typically are driven more miles. Pollution effects result from how much stuff cars emit and how many miles they are driven. So just targeting all old cars--and leaving newer cars alone--won't suffice.

It seems to be a relatively simple matter to measure E/M and M, and then to let the pollutor decide how to react: probably by some combination of decreasing E/G (regular tune-ups, for example), G/M (slowing down, maybe), or M.