First, I'll clear something up about Google's offering. It will actually load the smallest format your browser can handle. WOFF offers small file sizes, and your browser supports it, so it's the one you see. WOFF is also fairly widely supported. However, in Opera for example, you'll probably get the TrueType version of the font.
The file size logic is also, I believe, why Font Squirrel tries them in that order. But that is mostly speculation on my part.
If you're working in an environment where every request and byte counts, you'll have to do some profiling to find out which works best for your use case. Will people be only viewing one page, and never visiting again? If so, caching rules don't matter as much. If they're browsing or returning, Google might have better caching rules than your server. Is latency the bigger problem, or bandwidth? If latency, aim for fewer requests, so host it locally and combine files as much as possible. If bandwidth, go with whichever option ends up with the smallest code and smallest font format.
Now, on to the CSS vs JS consideration. Let's look at the following piece of HTML:
<head>
<script type="text/javascript" src="script1.js"></script>
<link rel="stylesheet" type="text/css" href="style1.css" />
<style type="text/css">
@import url(style2.css);
</style>
<script type="text/javascript">
(function() {
var wf = document.createElement('script');
wf.src = 'script2.js';
wf.type = 'text/javascript';
wf.async = 'true';
var s = document.getElementsByTagName('script')[0];
s.parentNode.insertBefore(wf, s);
})();
</script>
</head>
In many cases, script1, style1, and style2 would be blocking. This means the browser can't continue displaying the document until that resource has loaded (although modern browsers fudge this a bit). Which can actually be a good thing, especially with stylesheets. It prevents a flash of unstyled content, and it also prevents the giant shift that would occur when applying the styles (and shifting content is really annoying as a user).
On the other hand, script2 wouldn't be blocking. It can be loaded later, and the browser can move on to parsing and displaying the rest of the document. So that can be beneficial too.
Specifically talking about fonts (and even more specifically, Google's offering), I would probably stick with a CSS method (I like @import because it keeps styling with the stylesheet, but that could be just me). The JS file loaded by the script (http://ajax.googleapis.com/ajax/libs/webfont/1/webfont.js) is larger than the @font-face declaration, and just looks like a lot more work. And I don't believe loading the actual font itself (the WOFF or TTF) is blocking, so it shouldn't delay things too much. I'm not personally a huge fan of CDNs, but the fact is that they're REALLY fast. Google's servers will beat most shared hosting plans by a landslide, and because their fonts are so popular, people might even have them cached already.
And that's all I've got.
I have no experience with Typekit, so I left it out of my theorizing. If there's any inaccuracies, not counting generalizations between browsers for arguments sake, please point them out.
