Spring comes with various Servlet response wrappers that adapt specific headers to the HttpServletResponse, in particular to setContentLength and setCharacterEncoding where the Servlet response may hold a local setting that is not fully in synch with the corresponding HTTP headers (either in the Servlet container itself or in a Servlet response decorator applied by a Filter).

There is a key improvement that we can consistently apply: lazily and defensively parsing the content type, under the assumption that this adaptation code is usually not kicking in, therefore avoiding any unnecessary overhead in the common path. As a side topic, we can also use the new setCharacterEncoding(Charset) method on Servlet 6.1 now, avoiding unnecessary Charset->String->Charset transformations.