On the gauge-fixing problem