Huge peeve of mine: "non-semantic html" Who cares? Why bother? HTML isn't structured, extracting structured data (or the 'meaning'/'information'/'purpose') from soup is always going to be an inexact fumbling, even with ML or machine vision algorithms.

HTML 5 and it's deluge of new elements - used 100% correctly per the spec - is still only semi-structured. There are all kinds of ways to interpret what should go inside an 'article' or 'section' or 'nav' element and people do - you just can't trust the hundred million different interpretations of the spec every web developer brings to the mix. Then there is javascript which may make the page completely blank/unreadable if not enabled and, and .... urgh! Stop pretending like it's something that matters to you. It's only good for readability, I guess. Maybe. I'm not advocating going back to tables, but this is such a piffling little nothing to be concerned about. The important problem is can the information embedded in your webpage be easily extracted despite all the noise?

I don't know of any solutions to the tag soup that is HTML besides introducing structured data via microformats:

- https://support.google.com/webmasters/answer/99170?hl=en&ref...

- http://en.wikipedia.org/wiki/Microformats

- http://microformats.org/

And even that feels like a hack sometimes.

I've started writing a browser that sucks the information out of your webpages and displays it in a very boring, very Web 0.5 kind of way. It's going to have blackjack. And hookers.

RSS is the best thing since sliced bread.