+Old versions of curses, e.g., BSD 4.4, would return a null pointer
+from \fBinitscr\fP when an error is detected, rather than exiting.
+It is safe but redundant to check the return value of \fBinitscr\fP
+in X/Open Curses.
+.PP
+Calling \fBendwin\fP does not dispose of the memory allocated in \fBinitscr\fP
+or \fBnewterm\fP.
+Deleting a \fISCREEN\fP provides a way to do this:
+.bP
+X/Open Curses does not say what happens to \fI\%WINDOW\fPs when \fBdelscreen\fP
+\*(``frees storage associated with the \fISCREEN\fP\*(''
+nor does the SVr4 documentation help,
+adding that it should be called after \fBendwin\fP if a \fISCREEN\fP
+is no longer needed.
+.bP
+However, \fI\%WINDOW\fPs are implicitly associated with a \fISCREEN\fP.
+so that it is reasonable to expect \fBdelscreen\fP to deal with these.
+.bP
+SVr4 curses deletes the standard \fI\%WINDOW\fP structures
+\fBstdscr\fP and \fBcurscr\fP as well as a work area \fBnewscr\fP.
+SVr4 curses ignores other windows.
+.bP
+Since version 4.0 (1996),
+\fI\%ncurses\fP has maintained a list of all windows for each screen,
+using that information to delete those windows when \fBdelscreen\fP is called.
+.bP
+NetBSD copied this feature of \fI\%ncurses\fP in 2001.
+PDCurses follows the SVr4 model,
+deleting only the standard \fI\%WINDOW\fP structures.
+.SS "High-level versus Low-level"
+Different implementations may disagree regarding the level of some functions.
+For example, \fISCREEN\fP (returned by \fBnewterm\fP) and
+\fI\%TERMINAL\fP (returned by \fBsetupterm\fP(3X)) hold file descriptors for
+the output stream.
+If an application switches screens using \fBset_term\fR,
+or switches terminals using \fBset_curterm\fP(3X),
+applications which use the output file descriptor can have different
+behavior depending on which structure holds the corresponding descriptor.