The main features offered by phAtJAX are :

• Built-in debug console to show data transfer and processing as it happens, making debugging a snap!
• Totally server-side independent. No server code required to deploy. Obviously, server code is required to respond to the client, but it is not needed to deploy the UI. It can use any server-side language for data.
• Can handle streaming data (for browsers that can support it, i.e. Firefox).
• Options to send data via POST or GET. Send method can be configured on a call-by-call basis.
• Can receive data in 4 different data types: XML, string, name-value pairs and base64 encoded string. Data type can optionally be configured on a call-by-call basis.
• Data sent from server is automatically converted from given data type to a local JavaScript object that is immediately passed to your responder function for use.
• More than one instance of phAtJAX client can be loaded, since it is its own class instance.
• You can register calls to the server with the client so that you can manage them by checking status, aborting and getting a list of calls in the queue.
• Responder functions can be part of a local JavaScript class or just independent functions.
• Cache call data locally for common calls where the data rarely changes. Set cache time so that data is grabbed locally unless the cache time has expired, saving bandwidth and processing.
• Manage cache by having local cache automatically cleared if unused for a period of time.
• Validates server URL string using regex.

You can find sample code with phAtJAX available for download. Brian is also looking to offer more server side languages support in addition to PHP, which is a new orientation with AJAX frameworks to not be limited on one scripting language only.