if(ajaxSettings.dataType == ‘json’) {
remoteData = JSON.parse(remoteData);
}

is that it fails in Firefox <3.1 as it assumes native JSON. I looked at the way jQuery handle this in their own ajax function, and implemented it the same way (albeit with unsafe evaling of the JSON in the browser). The resulting modification is below:

if(ajaxSettings.dataType == 'json') {
remoteData = window.JSON && window.JSON.parse ?
window.JSON.parse( remoteData ) : (new Function("return " + remoteData))();
}

The trimming was required due to slop in the way xhr was implemented on some platform (trying to remember which). How trailing whitespace is dealt with is basically woefully inconsistent, so I trimmed for consistency.

I think this version of boostPeriod() is much clearer:

var boostPeriod = function() {
if (settings.multiplier > 1)
timerInterval = Math.max(timerInterval * settings.multiplier, settings.maxTimeout);
};

Constructing an optimized version of a function can be a useful technique, but here saving a property lookup and comparison once every 1 to 8 seconds is way overkill.

===

I’m curious, is it necessary to have success and error callbacks if you already have a complete callback registered? Complete should be called in either case, and is already checking the success status?

===

trimming each xhr and checking for ‘STOP_AJAX_CALLS’ seems quite app specific, it should at the minimum be a setting thats turned off by default, and documented.

Mike

Once you’ve got that, .stop() consists of cranking up the time to wait to something ridiculously high or setting a “stop” flag.