The price to pay with linears is poor efficiency. At high load currents, the pass transistor  in the regulator dissipates considerable energy, distributed as heat. This is where the  switcher is superlative. Here the pass transistor is switched on and off. Very little power is  wasted in heat due to very low "on" resistance and fast switching edges.

 Switchers are less reliable because they require more parts than the linear. Three pin linear  regulators are available, whereas the switcher needs a PWM, drivers, the switching  transistor(s), associated R and C networks, and magnetics.

 A nice attribute of switchers is isolation. A transformer can isolate the switching waveform  from the input d.c. bus. Rectifiers and filtering on the secondary convert back to d.c. The  linear regulator cannot be isolated. The output and input grounds are the same.

 Switchers, when operating with a transformer, can produce several voltages of either polarity.  This is why switchers are mostly used today for producing the various voltages needed in an  electronic project. Linear regulators can then be attached to low current switcher outputs to  improve noise characteristics, such as a quiet +/-15V. for op-amps.

 Switchers are a little fussier about input source impedance and output load impedance.  Switcher input impedance is normally negative, and the switcher can become unstable  (oscillates) if the source impedance (an L-C filter) does little to attenuate this negative  impedance. On the output, the main output voltage is normally fed back to the PWM module.  Output characteristics (poles and zeroes) become part of the overall loop gain. If they are  variable, then the loop gain could be affected, and oscillation could result. Linears, since  they are totally d.c., are less affected by I/O reactances.

 Switchers sometimes do not start at cold temperatures, due to insufficient loop gain of the  error amplifier. Newer PWM modules do not have this problem because the switching frequency  comes from an oscillator that always works. Regulation may suffer, however, at low temps.

 Switcher layout on p.c. boards is critical. Unwanted capacitive nodes of crossing or nearby  traces in high impedance circuits can affect loop gain. Parts associated with a PWM should be  grouped tightly nearby it. Usually, a breadboard may be tuned up and works fine, only to  discover a poorly laid out p.c. board causes oscillation of the regulator. High frequency  switchers should be laid out like r.f. amplifiers: short traces, spacing between sensitive  circuits, and ground planes for isolation, both on one side of the board and between layers.

 And finally, cost. As mentioned earlier, a linear regulator can be one part, a switcher many.  Thus, the linear is lower cost.