Question

In JSF 2.0, is there any significant technical difference between doing a non-AJAX submit e.g. and a “@all” AJAX submit e.g. ?

The user won’t perceive a page refresh from the AJAX-style submit, but is anything else technically significant happening? Is less data being submitted back to the server? Are any elements of the JSF life-cycle different with the AJAX versus non-AJAX submit?

Answer 1

Technically, @all is slower than a synchronous postback. There’s no difference in the HTML rendering during render response, but there is slightly more data in the ajax response, because the ajax response is been sent as a XML document with the updated HTML as a CDATA block. The following data is added to the response on top of the entire HTML output:

<?xml version='1.0' encoding='UTF-8'?>
<partial-response>
    <changes>
        <update id="javax.faces.ViewRoot">
            <![CDATA[
                HTML output here.
            ]]>
        </update>
        <update id="javax.faces.ViewState">
            <![CDATA[5778819104895950876:-4716773626508512118]]>
        </update>
    </changes>
</partial-response>

(you can see it yourself in “Net” or “Network” section of the web developer toolset in Chrome/Firebug/IE9 which you can get by pressing F12)

That’s thus always ~250 bytes more than a synchronous response. Also, there’s some overhead in the postprocessing because JavaScript has to parse all that HTML out of the XML response and replace the DOM with it, although the performance impact is practically totally negligible these days with fast machines.

However, in practice, @all appears visually faster than a synchronous postback because there’s no means of any “flash of content”.

Note that @all was unsupported for long in PrimeFaces because it is “fundamentally wrong” (to cite the PrimeFaces lead), but after the OmniFaces FullAjaxExceptionHandler, the PrimeFaces lead has changed minds and it’s supported from PrimeFaces 3.2 and on.