I always have just taken for granted that TLDs exist and are necessary to think about when you buy a domain name. But I'm wondering why they need to exist in the first place. I don't see why you couldn't have just had the domain name system be like this:
https://twitter/foo/bar https://example/foo/bar That is, I don't see a real need for there to be TLDs like .com and .org. It seems like an arbitrary thing. Am interested to know if there is any history behind this, or a need for it.
The organization of Domain Name System names has its origins in the growth of the internet from the old ARPANET. Details are contained in RFC 1034 from November 1987.
Before there was a DNS, all the name lookups were contained in a hosts.txt file that had to be distributed by FTP to all hosts. This does not scale well. The character of the systems on the network was also changing.
It was decided that a tree-based system would be adopted. This allows for distribution of name management, and for local networks to add internal names and systems without having to update external data.
Section 2.2 in the RFC describes the design goals of the system, which include a consistent name space; scalability; cost, speed, and accuracy trade-offs; applicability to a wide number of applications; independence from protocols and communications systems; and usability among the vast disparity of systems in use, from small personal computers to large mainframes and timeshare systems.
One other problem with not using a TLD is that in some OSs a single name can easily conflict with the name of another system on the local network. This has become less prevalent as more local networks use DNS internally, but there are still places that use the older local names (like the NetBIOS names used on Windows systems).
The biggest problem with this is that no two companies in the entire world could have the same domain name.
With the current system, the problem is limited to countries. While no two companies in Germany could have the same domain name, it is perfectly fine for a company in Germany to have the same domain name as a company in Austria, one in Switzerland, one in Spain, one in Australia, one in the US, one in New Zealand, one in Canada, and so on and so forth. (There are currently ~200 ccTLDs.)
Within a single country, the problem is much less prevalent. Usually, a company will want to use their trademark as a domain name, and trade marks are already unique within one market segment within one country.
And to add a bit more interesting history (having been involved personally pre www days (80's)):
ARPANET was a project of (D)ARPA - Defence Advanced Research Projects Agency which was focussed on developing reliable communications over disparate and unreliable data communications networks. This was the genesis of TCP/IP (IP "Internet Protocol" being the connectionless unreliable network layer) with TCP being the connection based, error correcting transport layer (Transmission Control Protocol).
Translating IP addresses into human readable names became important for usability, thus further up the network protocol stack came in DNS (Domain Name Service) that translated names to IP addresses (rather than relying only on the aforementioned hosts file on each machine).
In the beginning, there was really only .mil .edu and .gov TLDs. .com came along primarily for defence and government contractors. Most early usage was email, ftp and newsgroups (UUNET being another offshoot). gopher (gopher://) preceded http and was terminal text based hyperlinking. Basically http without graphics.
In Australia (where my involvement started), we had AARNET - Australian Academic and Research Network.
Separation of the domain name address space (and mapping to class A, B, C IP address ranges - 255.0.0.0 / 255.255.0.0 / 255.255.255.0) was an issue with ARAPNET (now beginning to be called the internet) spreading beyond its U.S. only birthplace.
So by various conventions (agreed through the RFC process) country abbreviations got added. So here we had .gov.au .edu.au .com.au. And to start with, that was about it.
A side note on RFC "Request for Comments" - was originally a very engineering focussed comment forum for further development of internet protocols. Its now become the standards body, with all the overwhelming formality and verbosity of any "standards" organisation. A quick read of the early RFCs is refreshing. For the really authoritative answer to this SE question have a read of: https://www.ietf.org/rfc/rfc1034.txt
So though the original motivation for dotted domain names was technical, certainly with the .com and .com.xx names, these have in later years become in embroiled in legalities of trade practices, commercial trademarks, et al, to the extent that its become impossible to both serve the technical / organisational objectives of a name space, and the commercial and geopolitical influences on the reservation of names (not to mention the parasites of domain name squatters who've made $billions over the years just for exploiting a dictionary and making it hard for people and organisations with real and actual purposes to have particular names).
So the coming of new TLDs is a welcome thing. It in effect deregulates some of the congestion that’s been tied up with lawyers and bureaucracies for a long time.
Why not in a single name space? (ie http://name/resource/resource).
Who owns the name?
And who owns the name in different countries which operate under different legal rules?
My surname could easily be McDonald......
As you say, there is no need for top-level domains (like .com and .org). But in the same sense, there is no need for domains at all (especially now given how search engines are integrated into browsers).
But domains are a useful organizational tool, as are the top level domains.
Note - top-level domains serve not ONLY to organize the namespace of names (not a very big deal for the examples you gave, but consider top level domains like .edu, and .uk etc for countries). And they also HELP to delegate responsibility for name assignment (UK government can dole out names for .uk domain, and another organization can dole out names in .us domain, and .com domain etc).
...And why they (almost never) host websites.
.UN, .EU)Like when accessing methods from a library, national ccTLDs denote namespaces. Unlike standard namespaces, these are demarcated by legal region, in accordance with ISO 3166 Alpha-2.
This means that a name can be present in a namespace whilst referring to a different organisation to one in another namespace. It permits organisations which exist in multiple countries with the same name, but are unrelated, to register the same domain in their relevant legal regions' ccTLD.
As an example, say that "Johnson Alan" in the USA and "Alan Johnson Limited" in the UK both want to register AJohnson.TLD. The former can choose .US, whereas the latter can choose .UK.
.org, .int)gTLDs exist because private organisations wish to run TLDs too, and an international private organisation may consider preferable to a nation, relying upon another private organisation that is dedicated to maintaining a TLD.
More exceptionally, gTLDs are useful because:
Their existence increases redundancy, by increasing the amount of root servers, thereby increasing the resiliency of the InterNet past what the amount of existent nations can provide.
ccTLDs may not be useful for individuals that are stateless, and organisations existing outside of national territory.
That is, I don't see a real need for there to be TLDs like
.comand.org. It seems like an arbitrary thing.
The fact that .com is so frequently utilized by non-companies, and .IO utilized for organisations registered outside of the British Indian Ocean Territory is ultimately failure of ICANN policy, but a deliberate decision by the TLD owners to prioritise profit and international control in the stead of organisation for the sake of altruism. It's not particularly unpredictable.
I don't see why you couldn't have just had the domain name system be like
schema://example.com/resource.
As serverfault.com/revisions/826234/1 explains, they are technically possible:
It is possible to create AAAA and A records on a TLD. There are a few TLDs which already do this and run a webserver on the address. Generally those will simply server a redirect to the website of the registrar for that TLD.
One upon a time, I had the undermentioned included in this answer:
However, were one to exist, you've formatted its URI incorrectly. Specifically, when not utilizing a TLD, you wouldn't be able to rely upon the TLDs routing you to ICANN's root nameserver, so you must include it. Consequently, the correct format for it would be
schema://domain./resource. The "." refers to ICANN's ultimate root nameserver, rendering the domain a "fully-qualified domain name" (FQDN).
I've since ascertained that that's untrue. If you're interested in why one would connect via an FQDN, I've asked software.codidact.com/posts/293325 to enlighten us.
Irrespectve, paraphrased (to remove now-incorrect information), webmasters.stackexchange.com/revisions/72697/6 states:
From a technical standpoint, it's possible, and
http://uz./is an example. It is a country code TLD, but the point is that DNS as a technology does support "dotless" domains.
This provides a way to differentiate these from hostnames, which are always scoped to the area network of the lookup origin.
Despite the standard theoretically and in practice permitting this, note that whether you'll be able to connect is another matter, per issues.chromium.org/issues/391903238#comment1. Specifically:
For me, firefox-134.0.1-1.fc41.x86_64 works:
...but google-chrome-canary-134.0.6972.0-1.x86_64 doesn't:
ERR_NAME_NOT_RESOLVED
However, ICANN doesn't want domains in its default namespace, as the previous source explains:
It appears that most generic TLDs are not allowed to have them due to ICANN polices. According to RFC 7085 such restrictions only apply to contracted TLDs, but it's not clear which TLDs fall under that restriction. Other sources suggest that the new TLDs would also be subject to such restrictions.
Equate this to how putting methods in the default C# namespace is very poor form, so they're always declared inside a namespace { method } encapsulator.
Theoretically, if your nation ever broke from ICANN, this would also render you unable to access your technically-TLD domain, whereas a gTLD would be able to work with your nation to provide access to its subdomains.
/posts/456302/revisions — I put a lot of time and effort into this, and it's all my own knowledge, except where explicitly aforecited. I suggest that you at least cite where I've supposedly plagiarised it from, else your criticism is unactionable. If that wasn't what you meant, my sincere apologies, and please rephrase your comment so that I might accurately respond. The biggest issue we would have without TLDs is mass limitations. In your examples, there could only be one "twitter" and one "example" website. This might be good in some ways, especially for large corporations fighting copyrights etc, but it would be a huge limitation to only allow one website for each word or combination of words.
The main idea initially was a good separation, like with ccTLD co.uk, .us etc, and .com was for commercial. But given the huge potential for re-sale, all of that seems to have been blown away in the name of making money (shame as it could have been a nicer system had ICANN enforced it in a more strict fashion).
While we don't even need domain names for the internet to work, and could just point out browsers at IP addresses, this too limits things as often a set of websites will be on the same server/network setup that has one external IP for many websites.
