getElementsByClassName, which lives up to its name by returning an array of DOM elements with a particular CSS class name.
But I don't need to concern myself with Rails or any of the scads of knock-offs based on Rails. Wiring prototype.js up to Zope required almost zero thought. Zope's been publishing objects and methods on simple URIs for longer than most systems have even existed, and throwing in a couple of Ajax specific view methods which performed their update and returned a simple string of "OK" (all that this application required) took no time at all. Note that I could very easily have dispatched off to a page template that returned HTML to display in line and Zope and prototype.js's Ajax commands would have been just as happy. So again, I tip my hat to Sam Stephenson and all who contributed for making it fairly platform agnostic when it comes to the server side. (note: the form serialization stuff could probably be tweaked / extended to work better with Zope's parameter marshalling).