triggers vs "NEW" pseudorecord

okay. postgres 7.0.3 here, on debian potato/stable.

from the docs at /usr/share/doc/postgresql-doc/user/c40874340.html
here is a working trigger:

CREATE TABLE emp (
    empname text,
    salary int4,
    last_date datetime,
    last_user name);

CREATE FUNCTION emp_stamp () RETURNS OPAQUE AS ' -- <= missing quote!
    BEGIN
        -- Check that empname and salary are given
        IF NEW.empname ISNULL THEN
            RAISE EXCEPTION ''empname cannot be NULL value'';
        END IF;
        IF NEW.salary ISNULL THEN
            RAISE EXCEPTION ''% cannot have NULL salary'', NEW.empname;
        END IF;

        -- Who works for us when she must pay for?
        IF NEW.salary < 0 THEN
            RAISE EXCEPTION ''% cannot have a negative salary'', NEW.empname;
        END IF;

        -- Remember who changed the payroll when
        NEW.last_date := ''now'';
        NEW.last_user := getpgusername();
        RETURN NEW;
    END;
' LANGUAGE 'plpgsql';

CREATE TRIGGER emp_stamp BEFORE INSERT OR UPDATE ON emp
    FOR EACH ROW EXECUTE PROCEDURE emp_stamp();

so altho the docs elsewhere say NOT to rely on access to the
pseudo table NEW within a trigger function, this part does work
like it should. but when i add SELECT or UPDATE it complains of
"NEW used in non-RULE query" -- what's the distinction?

what types of operations are NOT LEGAL within such a
trigger-invoked function? (i'd like to be able to UPDATE other
tables and SELECT from various tables within the function. bad
dog?)