I have to disagree almost entirely with this. Content is indeed separable from form in text (although form when applied to content significantly modifies perceived content -- as McLuhan says -- and form and content can be entwined in a way that's impossible to separate without losing information; but, your typical blog entry is not one of these circumstances). The thing is, in the context of the web, you cannot distinguish form from content effectively because of what amounts to a standardization error: you're really supposed to put your content in XML and your form in CSS, but XML bakes in a very important part of the form (where text is separated is part of the form, not part of the content) which CSS cannot then modify, meaning that CSS rearrangements of documents written with w3c best practices will be shallowly re-skinnable at best. For instance, if something is intended for the typical discussion forum form (wherein there is a title for the first post and then a sequence of posts following with no title of their own, single-thread), the XML representation cannot then be used purely in CSS to become (say) a blog post representation (wherein each post has a title, a part displayed on the main feed, a break, and then the rest of the post).
The solution is to consider plain unformatted text to be content and everything else to be form, and then to have a language for decorating, reformatting, and rearranging spans of plain text. In other words, drop embedded markup in favor of completely seperable markup (rather than pretending that embedded markup, because it's only used for blocking now, is no longer a part of the form and is instead content).
The result? You can take something written entirely in one form (Sherlock Holmes -- a serial) and automatically turn it into another form (a series of blog posts, complete with breaks) or overlay two unrelated text streams without changing the underlying content or bothering with xml parsing and rewriting.