To elaborate: In the very early days, HTML was often written by hand. Part of the early vision of the web was that anybody could write HTML and publish content.
It was thought that allowing malformed HTML to result in a completely blank page or an error message would be too unfriendly for less technical users.
In those early pre-CSS pre-JavaScript days, HTML was more like bbcode, Markdown, or RST today. It's interesting that AFAIK those languages don't have a notion of malformed input -- there is no movement to tighten the bolts on Markdown in the way that XHTML attempted to tighten the bolts on HTML.