* that requires the computation of a gradient, then the provided
* function is automatically projected along the curve, and the
* difference is only computed on the tangential part of the
- * gradient, since no information is available, on the finite
- * dimensional one, on the normal component of the gradient.
+ * gradient, since no information is available on the normal
+ * component of the gradient anyway.
* </ul>
*
* All functions use the finite element given to the DoFHandler object the last
* weighting function is given,
* i.e. weight=1 in the whole
* domain for all vector
- * components uniformly.
+ * components uniformly. Note that
+ * one often wants to compute the
+ * error in only one component of
+ * a solution vector (e.g. for the
+ * pressure in the Stokes system,
+ * when the solution vector also
+ * contains the velocity components).
+ * In these cases, the weight should
+ * be a <i>mask</i>, i.e., be a
+ * vector function for which individual
+ * components are either zero or one.
+ * This can easily be achieved using
+ * the ComponentSelectFunction
+ * class.
*
* It is assumed that the number
* of components of the function