Is nested XMLHttpRequests with multiple closures a good idea?

Question

I have a Greasemonkey script which operates on a search results page at a video site. The function of the script is to take a javascript link that opens a new window with a flash player, jump through some redirection hoops, and insert a regular link to the desired FLV file.

I have changed the script to do silly but structurally equivalent things to en.wikipedia.org. My question is whether the 3 nested closures and the nested xmlhttprequests is the best way to go about this.

---

// ==UserScript==
// @name                wiki mod example
// @namespace         http://
// @description         example script
// @include  *wikipedia.org*
// ==/UserScript==

var candidates = document.getElementsByTagName("a");

for (var cand = null, i = 0; (cand = candidates[i]); i++) {
  if (cand.href.match(/\/wiki\/W/)) { // for all articles starting with 'W'
    var progress = document.createElement('span');
    progress.appendChild(document.createTextNode(" Start"));
    cand.parentNode.insertBefore(progress, cand.nextSibling);
    progress.addEventListener("click",
    function(link1) { return function() { // link1 is cand.href
      this.innerHTML = " finding...";

      GM_xmlhttpRequest({method:"GET",url:link1,
        onload:function(p) { return function(responseDetails) {
          // p is is the current progress element
          // the first linked article starting with 'S' is *special*
          var link2 = responseDetails.responseText.match(/\/wiki\/S[^"]+/);
          if(!link2) { p.innerHTML = "failed in request 1"; return;}

          GM_xmlhttpRequest({method:"GET",url:"http://en.wikipedia.org"+link2[0],
            onload:function(p2) { return function(responseDetails) {
              // p2 is p, ie. progress
              // link3 would contain the URL to the FLV in the real script
              var link3 = responseDetails.responseHeaders.match(/Content-Length.+/);
              if(!link3) { p2.innerHTML = "failed in request 2"; return;}

              var elmNewContent = document.createElement('p');
              elmNewContent.appendChild(document.createTextNode(link3));
              p2.parentNode.insertBefore(elmNewContent, p2.nextSibling);
              p2.innerHTML = " <em>Done</em>";
            }}(p) // 3rd closure
          }); // end of second xmlhttprequest

        }}(this) // 2nd closure
      }); // end of first xmlhttprequest

    }}(cand.href), true); // 1st closure and end of addeventlistener
  } 
}

Answer 1

Well, you could improve readability by creating separate functions for each stage, then having stage 1 call stage 2, etc. So, instead of

request({onload: function(response) {
    request({onload: function(response) {
        request({onload: function(response) {
            alert("psych!");
        }});
    }});
}});

you'd have

request({onload: doTheNextThing});

function doTheNextThing(responseObject) {
    request({onload: doTheRightThing});
}

function doTheRightThing(responseObject) {
    request({onload: doTheLastThing});
}

function doTheLastThing(responseObject) {
   alert("psych!");
}

Answer 2

When this gets more complicated, you might consider a state machine. http://www.ibm.com/developerworks/library/wa-finitemach1/

Answer 3

Or if it gets even more complicated you can port Promises to your favorite JavaScript framework. AJAX programming is fundamentally broken. I've been doing it for 5 years - 40,000 lines of JS later this one attempt at a solution.