[djeglin]

its not the size of the image thats the only problem though... its the fact that for each image you're actually hitting the server again... each http request comes with a request/response header, and it takes the server some time to locate the file and serve it. If you're dealing with enterprise scale sites (Which you are obviously not) then this can make a big difference to load times when you have millions of page views per day.

In terms of the code... Yes, you can look at a number of "popular" websites and find horrible code. Just because it works doesn't mean its the right way to do things. I deal with semantic markup, my friend... Not botched code that "works".

Also, with your estimation that 99.9% of browsers will not have a problem with it, try thinking about what will happen if someone views the page CSS-naked. Yes, text-only browsers such as Lynx will ignore the images, but people with disabilities who strip out the css to make things more readable for them will still see the images, with no reason for them to need them. They will become a needless distraction. Also, to be properly compliant, all images need an alt tag - alt tagging your corner images will probably take as much markup as the nifty corners technique.

And finally, what about flexibility and maintainability? Using images, if you want to change the colour scheme of the site, you would need to create new corner images. Using a CSS technique, you simply need to change a couple of values in a CSS file that you would be modifying anyway.