JSONet uses regular external cascading style sheets (CSS) as the source of data. As the same domain restriction doesn't apply to linked stylesheets you can use JSONet to retrieve text data from any server besides yours.
Theoretically any browser anyhow supporting document.styleSheets is ready for
JSONet. That means any more-or-less descent modern browser.
Practically up to date the library was successfully tested on:
IXMLHTTPRequest / XMLHttpRequest based techniques ("AJAX") allow to seamlessly retrieve textual data and process it client-side. It is convenient, well supported, has a great number of libraries. But (and this is kind of "but" which may override everything else) it doesn't allow you to communicate with other domains; moreover it even doesn't allow you to communicate with different subdomains of your own server.
the cross-domain lock annoyance but it has other defaults.
The major one is that before you get on hold of data the received script will be automatically parsed and executed. That is like not knowing what kind of food is that before you eat it all (and wait for results).
A lesser known issue is that this technique has the US patent #6,941,562 pending. However unbelievable would look the application of this patent for royalties, the stories with GIF format and especially the recent multimedia case show that the most strange things may happen. While private users may not care, bigger companies may want to choose an Open Source solution instead.
To summarize JSONet has all advantages of the above mentioned techniques while not having any of their defaults.
Download (.zip with library and samples).
Sorry, CVS server is not configured yet, count the above as JSONet 1.0