Most lesser-known TLDs with a central registrar already do this by giving themselves the www.TLD. record; for example http://www.tv/ or http://www.name/. When you try just "tv./" or "name./" in your browser, it'll try those as well, without having to muddy up the main record.
http://canon./, on the other hand, refers to the "canon" that is located at the "root directory" of the internet—that is, ".". Most addresses don't need this, because once you specify any more than a host name, it assumes the rest of the domain "path" is a fully-qualified DNS record, and doesn't bother checking your local network for it.
This post automatically generated by Google.com in response to inarticulate mumblings by user anigbrowl
It has little value for users/consumers (i.e., won't remember if they want canon.com or canon.canon). It has little value for businesses, because it means more time and money monitoring for domain squatting and trademark infringement. It provides yet another playing field for scammers and other predators.
The only benefit is to the upper levels in the DNS hiearchy, gives them more revenue to perpetuate themselves and to come up with more user-hostile ideas.
Actually if you buy an arbitrary top level domain, there will be no more .com.
Instead it will be .google, .apple or .hackernews