When I was a certain corporate webmaster, I'd frequently be escalated because of a change in the markup for www·ibm·com. See, people would write applications which screenscraped something off www·ibm·com, perhaps the stock quote, our news headlines (which was perverse since we had a CDF feed from like 1997 on), whatever. People would screen–scrape and rely on the precise structure of the page, rather than request an API to the relevant content (likely because they couldn't justify the cost to us, or didn't have a legitimate reason to have access to machine-parseable content). Apparently this is happening more frequently as people write Greasemonkey scripts for Firefox. Just to repeat, a web page is not an API. It got to the point at IBM where we'd ban IP's we found doing machine harvesting of specific pages and republishing the content.

What people are doing with Greasemonkey is different, but somewhat the same: unless I as a webmaster explicitly say or make a covenant with you that a given URI is meant as an API (regardless of the format of the content), you're on your own. If I update it and it breaks your application: tough noogies. If you want to pay me to build a dependency on my application, then let's talk, but otherwise the idea that someone who provides content on the web is somehow obligated to make sure it works with random, arbitary applications is just bizarre. It was bad enough when we were expected that a given web page would work across every browser platform in creation.