Like jQuery and YUI, MooTools offers nice, browser-independent shorthand for manipulating arrays, divs, spans and whatnot. My favorite part continues to be the custom library construction tool that lets you select the functions you want. Check some boxes and get an entirely optimised version of MooTools with just the functions you need and none of the bloat you don't. That's lightweight.
A number of other libraries offer newer features, for animation or data visualisation or server-side processing or other niches, or different ways of thinking about life in the browser. To get a close look at some of these newer options, I unpacked a number of libraries, wrote a few lines of code, instantiated a few objects and pushed some code through a few browsers.
Animation and HTML5 game engines
One of the stated goals of HTML5, at least for some groups, is to replace the Flash plugin, the gold standard for making sprites and letters dance across the screen. This change is slowly coming as the game industry and the presentation industry start to duplicate some of the sophisticated tools available in the Flash universe.
Mashi is an impressive example of how the sprites can be set in motion. It offers more than several dozen standard easing functions for moving sprites along a timeline.
If you like board games, you'll like the three dimensional, orthographic views of game boards from pp3disco. It takes just a few lines of code to set them up. CasualJS was designed to be just like ActionScript for those with plenty of Flash experience. The authors at Crafty obviously spent a good amount of time on their collision detection. Fruit Ninja fans can play Crafty's Fruit Assassin in their browsers.
In theory, nothing could be easier than sticking a video tag into your HTML. In practice, the behaviour of so many supposedly standards-compliant browsers is different enough that you'll be pulling out your hair if you try to support all of them. A number of good HTML5 libraries let you write standard HTML5 video tags that will be replaced with Flash or QuickTime if the browser isn't ready to handle your video.
The Video for Everybody project offers one of the better libraries, filled with features that operate in the background to smooth delivery on older browsers like IE8. And if Video for Everybody lacks the features you need, you can check out spinoffs like VideoJS and MediaElement.js, two projects that began with the Video for Everybody code before adding their own layer of events and controls.
VideoJS, for instance, makes it easy to change the appearance with a skinning layer. Some of the CSS files contain no images and thus minimise bandwidth costs, at least aside from the expense of delivering the video file. MediaElement.js offers a number of meta features, such as controlling captions and subtitles. One example illustrates how to let people choose the language in which the subtitles appear.
Big data visualisation
If the pundits are correct, there are many, many terabytes of data just waiting for people to come along and try to make sense of the bits. The first generation of HTML5 libraries was more focused on building forms and generating tables. Now a number of libraries are zeroing in on building charts and drawing graphs on the object.
You'll find plenty of great options for basic line graphs, bar graphs and combinations. Flot, Flotr, Raphael and JSCharts are just a few of the libraries that deliver solid renditions of the classics.
Some libraries go even further. The collection of demos for Protovis and its newer cousin D3, a name meant as shorthand for Data Driven Documents, show how sophisticated constructions like Voronoi diagrams and network graphs can illustrate more than the up and down of some value. Simile Widgets offers a different collection of views that are more focused on maps and timelines.
These projects illustrate how we're just beginning to come up with good ways of turning data into pictures that can help us absorb large volumes of information quickly and efficiently.
One of the more interesting niches for visualising data is drawing network graphs on the tag, a feature that's useful for displaying social networks and flowcharts. Canviz is a collection of rendering algorithms that turn a network into a picture using a mixture of straight and curved lines.
Draw2D takes a different approach. It creates full Visio-like drawings that link together polygons to illustrate a workflow.