If the element has no children when we try to prepend a child, IE 10 (and probably other IEs) chokes on el.insertBefore(newNode, el.children[0]). Since el.children is an empty array (or nodelist, probably), accessing its first child returns undefined. MDN has a confusing paragraph about how undefined is treated differently than null in this case, yet things seem to work everywhere except IE.

Surely we can find a solution, possibly by checking the length of el.children first, and maybe explicitly passing null if need be, or... something.

It would be nice if addClass and removeClass could still accept a space-separated list of classes. We'll just handle the splitting and iterating. So to be clear:

• Detect browser support for el.classList
• Split the input string on spaces
• addClass
  • If classList is supported, iterate over each class and add it
  • If classList is not supported, just concat the entire string
• removeClass
  • If classList is supported, iterate over each class and remove
  • If classList is not supported, we still need to iterate over each class, removing it with a regex

Currently we're splitting name on . and treating everything as a classname. Really we should also:

• First split the name for [] pairs and push everything we find to the attrs object
• Also check for # in the name and use that to set an id
• Push all of the . classes we find to attrs so we can deal with all attributes at once
• Write some serious tests for all of this

If no element name is provided (name string begins with a symbol such as .), don't shift the nameParts, and instead set name to "div".

Many of our methods could expect a Node instead of an Element, particularly methods that insert children into the DOM. As it's currently written, our library won't accept the result of cloneNode() as a valid child, even though it totally is. This should be easy to fix without breaking any existing APIs.