Explaining how subpixel rendering technically works seems like the worst possible argument against something that exists purely for aesthetics. Just because something is more technically complicated does not make it a better option.

There are plenty of situations in which things are far less readable using subpixel rendering. Examples:

WITH subpixel: http://share.kyleneath.com/captures/Your_Dashboard_-_GitHub-... WITHOUT subpixel: http://share.kyleneath.com/captures/Your_Dashboard_-_GitHub-...

Furthermore, so long as the fonts are ~14px+ and sans-serif I think non-subpixel rendering makes the design cleaner and more aesthetically pleasing. Since I started slowly rolling it out to GitHub, I've been getting nothing but compliments. Case in point: http://github.com/training — I've had several people email me to ask me how I got the fonts to look so good on this page. The answer? The majority of the text uses antialiased fonts instead of subpixel.

Do I use this effect everywhere? Of course not. That's what being a good designer/developer being is about — surveying your options and choosing the one you feel is the best fit.

Now, you want to talk about font-rendering? Let's talk about monospaced fonts and Ubuntu for a minute because I've got a few rants down that road. Ubuntu's font-smoothing is nearing a crime against humanity. -webkit-font-smoothing doesn't have jack on it.