Jeff Croft wrote What's not to love about CSS frameworks, asking for debate over CSS frameworks like Blueprint, and got tons of it. Then he followed it up with some summary (and apology).

I don't do a whole lot of CSS work, but I've got some opinions anyway(!):

• Calling them frameworks may be a bit much. "Libraries" seems to fit the scope a bit better.
• Some designers complain of presentational class names. It's true that Blueprint encourages you to use class names like 'span-15'. The problem here is that CSS isn't rich enough to avoid it. It has no indirection capabilities. I'd like to be ables to say "style span-15 like this, and then style big-figure just like span-15", but I can't.
• People are starting to write tools to improve the expressiveness of CSS, like Moonfall, which provides variables and some simple margin math. Christian Montoya has a proof-of-concept of a tool called Semantify that lets you start with presentational Blueprint class names, and then fix them up to be pure semantic names.
• Some designers complain of inflexibility in Blueprint, but again, tools like Blueprint Grid CSS Generator help there too.

In the end, the biggest issues about the CSS frameworks seemed to be that many CSS designers simply prefer to do it all themselves. That's to be expected when sophisticated technologies are made simpler by providing simplifying libraries. The experts have climbed the learning curve, and can do more with the raw technology, they find the framework limiting. Non-experts find the raw technology baffling, and appreciate that the framework simplifies and organizes their work.