For example how would the Google Chrome browser work (and search to a degree), does a user who types "pepsi" into their omnibar want to go to http://pepsi/ or search for pepsi?
It also adds confusion for webmasters trying to parses URLs on their website. When a user types pepsi.com/coke into their blog for example that link can be made clickable to http://pepsi.com/coke/ where it can't for pepsi/coke
The problem of detecting URLs with weird TLDs could be solved by using relative URLs (RFC 1808 §4) with the protocol omitted. It would put that vestigial // to good use—//coca-cola is a URL just as @cocacola is a Twitter handle.
For example how would the Google Chrome browser work (and search to a degree), does a user who types "pepsi" into their omnibar want to go to http://pepsi/ or search for pepsi?
It also adds confusion for webmasters trying to parses URLs on their website. When a user types pepsi.com/coke into their blog for example that link can be made clickable to http://pepsi.com/coke/ where it can't for pepsi/coke