Index: Doc/lib/libfuncs.tex
===================================================================
--- Doc/lib/libfuncs.tex	(revision 50627)
+++ Doc/lib/libfuncs.tex	(working copy)
@@ -477,6 +477,24 @@
   \code{[item for item in \var{list} if function(item)]} if function is
   not \code{None} and \code{[item for item in \var{list} if item]} if
   function is \code{None}.
+  
+  When dealing with instances of subclasses of \class{str}, \class{unicode}
+  or \class{tuple}, \function{filter()} utilises a number of shortcuts for
+  performance reasons:
+  
+  \begin{itemize}
+    \item \function{filter()} will not invoke any overridden
+          \method{__len__()} methods;  that is, it will explicitly invoke
+          e.g. \code{str.__len__()}.
+    \item \function{filter} will not utilise any implementation of the
+          iterator protocol defined by the subclass.
+  \end{itemize}
+  
+  Further note that in spite of these points, \function{filter()} \emph{will}
+  call the subclass's \method{__getitem__()} method.  To sum up, when
+  fetching the object's elements, \function{filter()} will use a
+  combination of the subclass's \method{__getitem()__} and the parent
+  class's \method{__len__()}.
 \end{funcdesc}
 
 \begin{funcdesc}{float}{\optional{x}}