I'm not sure what TLDs have to do with slashes or vise versa.
http://to/ is using only a TLD, it's not that it doesn't have a TLD.
Many of the URL components are optional and if not specified designate relative URLs to be filled in with parts from the base URL. For example, if all your internal URLs in a document are //domain/whatever, that is they don't specify a protocol, then the page is completely viewable on both HTTP and HTTPS protocols without having the browser indicate mixed security contexts. Without the double leading slashes indicating the hostname, you'd be unable to recognize that a URL is relative to the protocol vs relative to the dirname of the current document. In the same way, http:/file fills in the hostname from the current base URL.
TLDs don't really mean much technically, they were created for human consumption. There may be load balancing issues that are made easier because of TLDs, but it's not technically a requirement. On your own name server, you can create a zone that ends with anything (try it).
There have been recurring attempts/suggestions to get rid of TLDs, somewhat as a response to the needlessly ever-expanding TLD list, and somewhat as a response to break up the monopoly that exists in determining them ("If only more people could make their own TLDs, the world would be a better place, woes is us!"). More TLDs mean more costly trademark enforcement and continual attempts to corral different kinds of services/websites into predictable naming schemes (.xxx, .sex, .travel, .biz, and the super-stupid .tel, .name, and .mobile) -- people have specifically been against .xxx or .sex TLDs with the argument that it would be easier to convince law/policy makers to force people to use them for adult sites and thus limit free speech, for one example.
I don't really mind TLDs, I'm not convinced that lack of TLDs would be any kind of significant win. In the same vein, I don't like San Francisco's apparently random street numbering scheme and fake grid system (random streets that run for a block start at 0, no matter what the surrounding addresses are, or a single block is bounded by different address ranges (check out the block that Sanchez Elementary School is on http://maps.google.com/?ie=UTF8&hq=&hnear=San+Franci... this block is bounded by the 300 and 400 on the west and east, and 3400 and 3700 on the north and south), which disturbs this boy's sensibilities being a nerd who grew up in Chicago, which is on a strict grid system), but I'm not going to suggest that it change either.
I realize http://to/ is a TLD and slashes don't really have much to do with TLDs; they just got me thinking about URLs in general. Relative URLs with the protocol specified is something that hadn't occurred to me before. Thanks!
One question- I'm not too sure how a DNS lookup works from start to finish, but creating your own arbitrary zones wouldn't work for most people, would it?
Well, you'd do that for internal purposes. Anyone could set one up, so in that sense it would "work" for most people, but it wouldn't resolve for people who don't use the name server that the zone is setup on. I can see a use for it, but I think it's easier, conceptually, to create a private subdomain within an already established and registered publicly accessible domain and limit its access with views (a capability/feature of ISC bind (and perhaps others) where different zones are made available based on the source address of the DNS query).
The .local TLD, which is where zeroconf/rendezvous hosts end up is a standardized example of that. It is resolved not via regular name servers, but rather a network-segment limited distributed broadcast (an oversimplification), however.