How much do you remember from elementary school? I remember vinyl tile floors, the playground, the teacher sentencing me to standing in the hallway. I had a teacher who was a chess fanatic; he painted a huge chess board in the paved schoolyard and got someone to fabricate big wooden chess pieces. It was enough of an event to get us on the evening news. I remember Run for the Arts, where I tried to talk people into donating money on the theory that I could run, which I could not. I'm about six months into trying to change that and I'm good for a mediocre 5k now, but I don't think that's going to shift the balance on K-12 art funding.
I also remember a domain name: bridger.pps.k12.or.us
I have quipped before that computer science is a field mostly concerned with assigning numbers to things, which is true, but it only takes us so far. Computer scientists also like to organize those numbers into structures, and one of their favorites has always been the tree. The development of wide-area computer networking surfaced a whole set of problems around naming or addressing computer systems that belong to organizations. A wide-area network consists of a set of institutions that manage their own affairs. Each of those institutions may be made up of departments that manage their own affairs. A tree seemed a natural fit. Even the "low level" IP addresses, in the days of "classful" addressing, were a straightforward hierarchy: each dot separated a different level of the tree, a different step in an organizational hierarchy.
The first large computer networks, including those that would become the Internet, initially relied on manually building lists of machines by name. By the time the Domain Name System was developed, this had already become cumbersome. The rapid growth of the internet was hard to keep up with, and besides, why did any one central entity---Jon Postel or whoever---even care about the names of all of the computers at Georgia Tech? Like IP addressing, DNS was designed as a hierarchy with delegated control. A registrant obtains a name in the hierarchy, say gatech.edu, and everything "under" that name is within the control, and responsibility, of the registrant. This arrangement is convenient for both the DNS administrator, which was a single organization even after the days of Postel, and for registrants.
We still use the same approach today... mostly. The meanings of levels of the hierarchy have ossified. Technically speaking, the top of the DNS tree, the DNS root, is a null label referenced by a trailing dot. It's analogous to the '/' at the beginning of POSIX file paths. "gatech.edu" really should be written as "gatech.edu." to make it absolute rather than relative, but since resolution of relative URLs almost always recurses to the top of the tree, the trailing dot is "optional" enough that it is now almost always omitted. The analogy to POSIX file paths raises an interesting point: domain names are backwards. The 'root' is at the end, rather than at the beginning, or in other words, they run from least significant to most significant, rather than most significant to least significant. That's just... one of those things, you know? In the early days one wasn't obviously better than the other, people wrote hierarchies out both ways, and as the dust settled the left-to-right convention mostly prevailed but right-to-left hung around in some protocols. If you've ever dealt with endianness, this is just one of those things about computers that you have to accept: we cannot agree on which way around to write things.
Anyway, the analogy to file paths also illustrates the way that DNS has ossified. The highest "real" or non-root component of a domain name is called the top-level domain or TLD, while the component below it is called a second-level domain. In the US, it was long the case that top-level domains were fixed while second-level domains were available for registration. There have always been exceptions in other countries and our modern proliferation of TLDs has changed this somewhat, but it's still pretty much true. When you look at "gatech.edu" you know that "edu" is just a fixed name in the hierarchy, used to organize domain names by organization type, while "gatech" is a name that belongs to a registrant.
Under the second-level name, things get a little vague. We are all familiar with the third-level name "www," which emerged as a convention for web servers and became a practical requirement. Web servers having the name "www" under an organization's domain was such a norm for so many years that hosting a webpage directly at a second-level name came to be called a "naked domain" and had some caveats and complications.
Other than www, though, there are few to no standards for the use of third-level and below names. Larger organizations are more likely to use third-level names for departments, infrastructure operators often have complex hierarchies of names for their equipment, and enterprises the world 'round name their load-balanced webservers "www2," "www3" and up. If you think about it, this situation seems like kind of a failure of the original concept of DNS... we do use the hierarchy, but for the most part it is not intended for human consumption. Users are only expected to remember two names, one of which is a TLD that comes from a relatively constrained set.
The issue is more interesting when we consider geography. For a very long time, TLDs have been split into two categories: global TLDs, or gTLDs, and country-code TLDs, or ccTLDs. ccTLDs reflect the ISO country codes of each country, and are intended for use by those countries, while gTLDs are arbitrary and reflect the fact that DNS was designed in the US. The ".gov" gTLD, for example, is for use by the US government, while the UK is stuck with ".gov.uk". This does seem unfair but it's now very much cemented into the system: for the large part, US entities use gTLDs, while entities in other countries use names under their respective ccTLDs. The ".us" ccTLD exists just as much as all the others, but is obscure enough that my choice to put my personal website under .us (not an ideological decision but simply a result of where a nice form of my name was available) sometimes gets my email address rejected.
Also, a common typo for ".us" is ".su" and that's geopolitically amusing. .su is of course the ccTLD for the Soviet Union, which no longer exists, but the ccTLD lives on in a limited way because it became Structurally Important and difficult to remove, as names and addresses tend to do.
... continue reading