DNS is hierarchical. You know those dots? They actually mean something. Each name is a separate layer you have to query. Each one means more work for people who maintain infrastructure like the public suffix list[1]. Sure, each one might not count for much, but a world with thousands of TLDs isn't better for anyone.
Besides: TLDs function as a lexical tagging mechanism. When you see a string of letters like "foo.com", you recognize the word as a domain name because it's tagged with ".com". If we have thousands of TLDs, it's harder to recognize domain names as domain names.
I don't think any of these concerns will stop our descent to the world of unstructured AOL-keywords-as-DNS, though, and that's a shame.
Domain names are a mess. They're backwards (should be us.ag.bobsforestryservice), and since most domains just end with .com, there's no hierarchy. In my opinion, if ICANN wanted a hierarchy, they should be strictly regulating the names, including checking ownership of trademark, checking country of registration, etc. So thus we have AOL keywords.
That's the second time I've seen this argument made on HN this week. Why does that make more sense?
If I start typing in my address bar "g" then google.com is the first completion suggestion. If I want to get more specific, I can type "n" and news.google.com becomes the first suggestion. Or "m" and I go to mail.google.com. If we had to write "com." before any of these came up, it would mean a serious loss of productivity for everyone. The current system provides one or two layers of specificity before the TLD, then as many more as you would like (like /item?id=9357898 on the end of this URL, something nobody will every type) - it's ideal for everyday use even if it doesn't fit into some clean sorting method you're imagining or something.
Because directory hierarchies go from top to bottom - /dir/subdir/file.ext - or, globally, //hostname/dir/subdir/file.ext, or protocol://hostname/dir/subdir/file.ext etc.
That the hostname part's components, as presented to users, goes from bottom-to-top in DNS when the rest goes from top-to-bottom is an accident of history, but one it's too late to change (in DNS). Not everything made that mistake however - Usenet didn't.
As for what you're typing in your 'awesome bar', when you start typing, your autocorrect is ranking your visited history: there's no reason it has to start at the beginning, especially when the beginning isn't the root, but there's also no reason that doesn't make sense.
In fact, drifting back to topic: GOOGLE. is in fact a TLD now. If DNS were the 'right' way round, you'd be going to //google.news - wouldn't that make more semantic sense?
Of course, in practice, we're stuck with DNS the way it is because .com is now firmly in the public consciousness. But it could easily have been different, and if I were designing something new, I'd pick the Usenet way round.
Autocomplete could work the other way around, with "n" finding `∗.news` and `g.news` finding `∗g.news`, after all that `news.google.com` actually starts with `http://`.
Yes, and I got downvoted by this crowd last time I mentioned as much. Some random SF-specific project launches, and snags up yet another .com domain. It's not even something like "san_fran_project.com", it was even more generic.
On that note, I'm fully with the GP. This sort of thing should be enforced. It would help with everything from spam, to fake sites, to weird domain name pollution as discussed here.
They should really have done that. TLDs that are properly regulated do have a meaning and are successfully used in their space. Examples are .gov or .edu (except that they should have put in a country layer) and .de, which is the most successful CTLD, because unlike .co or .io it is used only by entities that actually have a physical address in Germany.
Besides: TLDs function as a lexical tagging mechanism. When you see a string of letters like "foo.com", you recognize the word as a domain name because it's tagged with ".com". If we have thousands of TLDs, it's harder to recognize domain names as domain names.
I don't think any of these concerns will stop our descent to the world of unstructured AOL-keywords-as-DNS, though, and that's a shame.
[1] https://publicsuffix.org/list/effective_tld_names.dat