Why don't we just use element IDs as identifiers in JavaScript?

All browsers I've come to work with allow accessing an element with id="myDiv" by simply writing:

myDiv

See here: http://jsfiddle.net/L91q54Lt/

Anyway, this method seems to be quite poorly documented, and in fact, the sources I come across don't even give it a mention and instead assume that one would use

document.getElementById("myDiv")

or maybe

document.querySelector("#myDiv")

to access a DOM element even when its ID is known in advance (i.e. not calculated at runtime). I can tell that the latter approaches have the advantage of keeping the code safe if someone inadvertedly attempts to redefine myDiv in a wider scope (not such a brilliant idea though...), overwrites it with some different value and goes on without noticing the clash.

But other that that? Are there any concerns in using the short form above other than code design, or what else am I missing here?

Anyway, this method seems to be quite poorly documented, and In fact, the sources I come across don't even give it a mention [...]

Reliance on implicitly-declared global variables aside, the lack of documentation is a great reason not to use it.

The apparent promotion of id values into global variables isn't standards compliant (the HTML5 spec for the ID attribute doesn't mention it) and, therefore, you shouldn't assume future browsers will implement it.

EDIT: It turns out this behaviour is standards compliant - In HTML5, window should support property access to "Named Elements":

Named objects with the name name, for the purposes of the above algorithm, are those that are either:

• child browsing contexts of the active document whose name is name,
• a, applet, area, embed, form, frameset, img, or object elements that have a name content attribute whose value is name, or
• HTML elements that have an id content attribute whose value is name.

Source: HTML 5 spec, 'Named access on window object", emphasis mine.

Based on this, standards compliance is not a reason to avoid this pattern. However, the spec itself advises against its use:

As a general rule, relying on this will lead to brittle code. Which IDs end up mapping to this API can vary over time, as new features are added to the Web platform, for example. Instead of this, use document.getElementById() or document.querySelector().

Great question. As Einstein probably didn’t say, things should be as simple as possible, and no simpler.

the latter approaches have the advantage of keeping the code safe if someone inadvertedly attempts to redefine myDiv in a wider scope (not such a brilliant idea though...), overwrites it with some different value and goes on without noticing the clash

That’s the main reason why this is a bad idea, and it’s quite enough. Global variables aren’t safe to rely on. They can be overwritten at any time, by any script that ends up running on the page.

In addition to that, just typing in myDiv isn’t a “short form” of document.getElementById(). It’s a reference to a global variable.document.getElementById() will happily return null if the element doesn’t exist, whilst attempting to access a non-existent global variable will throw a reference error, so you’d need to wrap your references to the global in a try/catch block to be safe.

This is one reason why jQuery is so popular: if you do $("#myDiv").remove(), and there is no element with an id of myDiv, no error will be thrown — the code will just silently do nothing, which is often exactly what you want when doing DOM manipulation.