@tehyoyo, border radius is a huge task for browsers and there is much doubt in my head as to whether an image would take less processing time than border radius :). Both of which don’t really matter because of the speed these days 
NOT to drag things way off topic, but…
Most of the ‘debate’ is by people who don’t take time to think about it or understand that shock screen and print are two radically different devices (though Apple is trying to do something about that with ‘retina’)
Serifs DO help legibility in print… there are tons of studies that show it makes it easier to distinguish each glyph when the glyphs are fully formed… but to be fully formed you need analog print (daisy wheel, old-fashioned typeset) or really high DPI (300+)… this is consistent with modern print where your average professional printing STARTS at 1200dpi. That’s plenty of dots in a inch to form the entire glyph and it’s serifs properly. a 12 point character getting anywhere from 50 dots height on a crappy 300dpi laser to as much as 800 dots in high end professional lithography.
Screen just doesn’t have the resolution or pixels per inch for that… you take a 14px “comfortable” font from a OS that ‘thinks’ it’s running at 96dpi (which is usually a lie on the actual display) that gets if you are lucky 13px from the caps-line to the descender… roughly a quarter of normal print from a cheap printer and half that of the barely legible ‘draft’ print. There aren’t enough pixels to render serifs as anything more than a blurry mess… which is why serifs on screen are usually a bad idea at ‘normal’ (12px to 18px) sizes. You bump the sizes up to give them enough pixels and combined with anti-aliased/font hinting rendering, then you can use serifs.
It’s also why you never saw a whole lot of serif fonts being used in the 9 pin dot matrix printer days.
Though there ARE exceptions – and we don’t get to use them. Raster fonts can often get away with serifs on as little as 8px height – because each character is crafted EXACTLY for that size. Dynamic/vector type fonts – like postscript and truetype – can’t be done that way. The original IBM PC CGA text is a great example of this – a legible serif font that fits each character into 8x8 with no hinting/smoothing/aliasing. Good luck getting that out of a vector font.
It’s something the people who pimp vector graphics like they were the second coming don’t like to talk about. Vectors are great for scaling up – but scaling down below a certain point they fall apart miserably; and that’s where raster graphics really start to shine… though raster of course falls apart the moment you try to resize it.