A Bit About DNS, HTTPS, and DNS Over HTTPS

Globe

Browsing the web involves lots of different protocols that most people probably don’t consider.

DNS is one of those protocols, so in this post, we’ll briefly go over the basics of DNS, why browsers are switching to DNS Over HTTPS (DoH), and how to get public DNS Records.

What is DNS

The core function of DNS (Domain Name System) is to associate Domain Names to IP Addresses and vice versa. The concept is deceivingly simple, and incredibly useful.

DNS is commonly used as an abstraction for complex web applications that would otherwise be very hard to manage and scale. It has many features that can be discovered by the types of “Records” that DNS keeps.

DNS Records

Information about domains are stored in “Records” that have different things they can point to, such as different IPs or other domains.

Here are some common records for example:

The Application Layer

Much like HTTP and HTTPS, DNS relies on underlying protocols like IP for a Addressing and TCP/UDP for data transfer. These lower layer protocols describe how computers get addressed, how data gets segmented, and the manner in which it gets sent over the network. But the rest is up to the application or OS to figure out. Thankfully standard protocols have emerged and evolved, driven by committees, companies, and people with a vested interest in the technology.

Allowing more and more clients to communicate with servers more efficiently is great, but often requires a collaborative effort with everyone who makes devices, applications, libraries, OS’s and servers on the internet. The history of HTTP(s) is a good place to see this from HTTP/1.1 to HTTP/2 and HTTP/3. From SSL to TLS to encrypt web traffic. Now we’re seeing DNS come over an encrypted channel, which is a good thing for privacy if done right.

Protocols and Ports

The protocols mentioned so far are considered standard and have lower port numbers. The first 1024 ports are usually reserved by the OS specifically for these protocols. For example some common ones are:

  PROTO     PORT    NAME
  ----------------------
  TCP/UDP   22      SSH
  UDP       56      DNS
  TCP       68-69   DHCP (Dynamic Host Configuration Protocol)
  TCP       80      HTTP
  TCP       443     HTTPS
  TCP       123     NTP (Network Time Protocol)

Using netstat -tan or ss -tan, you can see all the ports that your computer is listening, or sending data to. These commands actually show open sockets, but that’s another topic.

Problem Domain

Even in the 70’s, when the internet was in its infancy, people realized that computers were much better at remembering numbers than humans, and early systems were created for mapping “hostnames” to IP Addresses. The term “hostname” and “hosts” is still used today, but this is not to be confused with a “Domain Name” which is a similar idea, but can be entirely different identifiers.

Hosts and Hostnames

On Linux and Unix based systems, there is a file called /etc/hosts, which is a plain text configuration file containing matching ip addresses and hostnames. This effectively makes a hostname synonymous to an IP. There is also a command hostname and a file called /etc/hostname, which contains – you guessed it – your hostname. On windows, there is a hostname command too, but the hosts file is located at c:\windows\system32\drivers\etc\hosts.

Blocking Hosts

An easy way to “block” a website is to use 0.0.0.0 for a hostname in your hosts file.

This project called hosts using Python to curate hosts files to block malicious domains.

PiHole is popular project for ad-blocking at the network level with DNS. It does this by spoofing a DNS resolver and using a blacklist that blocks known ad-domains. The setup is easy and automated and it runs on a Raspberry Pi and has Docker images.

Setting Client DNS

You can configure what DNS resolver your computer uses, but the process will be different depending on your OS. On a Linux distro using NetworkManager, a file called /etc/resolv.conf points to the DNS Server which will resolve your lookups.

If your router provides DNS, it’s probably the default for your computer. Like an IP address, routers typically get their DNS settings automatically from an ISP. Of course you don’t have to use your ISP’s DNS servers, you’re free to use any server you like.

Public DNS resolvers

Cloudflare hosts a public DNS server at 1.1.1.1 and claims to be “the world’s fastest DNS” and Google hosts their public dns server at 8.8.8.8. Here’s a comprehensive list of public DNS resolvers

Domains and Domain Names

So Domain Names and DNS are a lot like the local hosts file on your computer, except as a global distributed server that has a list of all registered domains and their IPs.

Full Domain Names have a lot more restrictions than hostnames and are generally composed of several different fields separated by a . (dot). The dot notation is tied to the recursive way that DNS resolvers search for domains. If you don’t know what that means, it’s ok, we will get to that later.

A domain name is essentially a public record of a hostname, so you must make an account with a registrar who can reserve the domain on your behalf. The price of a domain can range from $5 to tens of millions and it all depends on scarcity and demand.

A TLD (Top Level Domain) is the last (rightmost) part of a domain. The most common in the US being .com that most American companies use. Other common TLDs include .org .net .edu .gov. There are many many more TLDs, with specific purposes, like ones for a specific countries (.uk .jp .fr .au), for specific services (.travel .jobs .bike) and even specific businesses (.bmw .bosch .nike).

A full list can be found here. IANA is the Internet Assigned Numbers Authority, and they are the organization that gives your ISP a block of IP Addresses. They also own the “Root Zone Database” for DNS Records.

TLDs specify the DNS Root Zone of a FQDN (Fully Qualified Domain Name), but what about the domain itself?

Subdomains

What comes immediately to the left of the TLD is a “child” domain name. For example there can be a second level domain such as .co.uk. This is where the “recursive” part of DNS happens as soon as you start adding subdomains. The . is the one and only delimiter that separates subdomains, so domains are nice, clean and uniform, unlike some URLs.

All TLDs have at least one subdomain and there is no uniform specification for how subdomains should be set up, however, there are best practices. Some examples are en.wikipedia.org mail.google.com www.github.com.

Naming subdomains appropriately helps people determine their purpose. Taking advantage of the recursive nature of DNS by setting up subdomains can make operations much easier and more organized. It can save organizations from registering n new domain for every new service.

URLS

Another distinction that often confuses people is the difference between a URL and a Domain Name. Browsers do their best to figure it out, but sometimes this doesn’t work. For example browsers automatically assume ‘https://’ as the scheme of the URL and port 443 or 80 as the default port, so this leaves the user to type in – at minimum – the domain name, and the ending part of the URL, often refereed to as the ‘slug’ or ‘path’ is optional.

Although a URL may seem like a basic thing to use, it can be surprisingly tricky to make. It is interesting to see what characters it does and does not allow and why.

To give a run down of the syntax from the URL Wikipedia Page

  URI = scheme:[//authority]path[?query][#fragment]
  authority = [userinfo@]host[:port]

Here we see after a scheme:, the [//authority] is often shortened to a //host, or the DNS equivalent of a host – a Domain Name. From there, a path is required. If no path is provided, a browser will look for an “index.html” file in the root directory of the web server.

URLs often require special encoding and escaping, particularly in the case of accepting user input and sending it as a query string. There’s lots more information about safely encoding a URL on the web, so I’m not going to go into that right now, but needless to say URLs can become quite unruly and should be constructed very carefully.

Clean URLs

Having so called “clean urls” is often desired for web sites. This method of “Rewriting” the url can be configured in either the web server, or the web framework in the code where URLs of Requests and Responses are handled.

You will notice that the majority of web sites these days do this with their URLs. It is a good practice to not include the name of the file, and only put information that is absolutely necessary in the URL.

Not only does it help with SEO and web crawlers, but it also helps humans who may need to work with those URLs.

Slugs vs Subdomains

The decision to use URL Slugs to Subdomains often depends on the scope and size of an organization or website in question. Something to consider is how often and how likely the name is to be changed in the future. Changing a subdomain can take a long time to propagate while changing a URL Slug on a server is instant. Generally, applications will cache DNS lookups for a long time as well, but this too can be configured.

Wikipedia could have used wikipedia.org/en/ for example, but it is unlikely that the English version of Wikipedia will go away or change, so I think a subdomain here is the right choice. Articles on the other hand are very likely to change, at the server level, so it makes sense to have them in the URL Slug. I think generally people are better at remembering Subdomains than URL Slugs, but I’m really just speaking for myself.

APIs

A pretty big grey area where you see lots of variance between slugs and subdomains is with Web APIs. Generally, public APIs used by companies will have their own subdomain like api.github.com or api.paypal.com/v2/ api.spotify.com/v1/. This is a very common scheme, but there’s also endpoints that use slugs like example.com/api/v1 or just a plain JSON file that can be requested instead of html like example.com/data.json.

WWW

Since we’re on the topic of Domains, URLs and Slugs, it is only appropriate to mention the www. subdomain prefix, why it matters and why people still use it.

The www. was traditionally used to indicate what type of server the domain was for, if there was any doubt whether it was ftp or irc server, you could tell by the domain.

This was never a rule and was never enforced, but has become very common practice over the years. There is virtually no downside to creating a CNAME entry for www.yoursite.com that just points back to an A Record (yoursite.com). Another method is to use HTTP Redirects.

The best thing to do is handle and test both cases, because both ways are still pretty common, and it would be a shame if somebody went to www.yoursite.com and they get a fat ERR_NAME_NOT_RESOLVED message, because you forgot to redirect that traffic back to your domain.

This Netlify article goes over some of the reasons to use www. or not.

Setting DNS For A Server

Registering a Domain Name is often the first step in creating a public website. This is done at a domain name registrar like namecheap or Google Domains.

After you buy a domain, you have full rights to do whatever you like with it. You can redirect it to any IP (even someone else’s), use it as load balancer, use it as a cache, set up DDoS protection. If you’re hosting a website, you will set the A Record to point to the IP that hosts your website.

DNS-Over-HTTPS

Traditional DNS is a plain text protocol that makes it easy for someone to snoop on a network and see all the domain lookups.

The solution purposed for this is to move DNS over an encrypted protocol, such as HTTPS. The RFC has been around since 2018, and Firefox kicked off this feature in browsers last year. Here is the article about it.

Firefox

As of February 2020, Firefox’s default setting ignores the DNS settings of your OS and uses their own.

My opinion is that intentionally disregarding a user’s network settings is a pretty intrusive, even if in the name of “security” and “privacy”.

Ultimately, I chalk this up to a hasty and lazy implementation of a new protocol, but I hope they come back to respecting their user’s system preferences soon.

I went back to see how Chrome and Chromium do DoH, and this is what I found.

Chrome

As of writing, Chrome’s adoption of DoH has been slower, but it seems they are taking a much more reasonable approach.

This page explains how Chrome will try to “upgrade the protocol used for DNS resolution while keeping the user’s DNS provider unchanged

More Centralized DNS

Something to think about is if a critical mass of people switch to the same resolver, it can be harmful to the distributed nature of DNS.

Think about Chrome’s near 70% share in the browser market. If they did what Firefox did, and updated Chrome to use Google’s DNS by default, it would be a massive blow to every other DNS provider out there and raise eyebrows.

DNS Utils

There are more interesting ways that we can poke at DNS to find out more about a certain domain.

Here’s some CLI tools to get info on hosts and domains.

getent

Gets an entry from NSS (Name Switch Service) Libraries. Almost guaranteed to be on any GNU/Linux system with a network stack. getent hosts is basically the command equivalent of calling the gethostbyname() function in a C program.

NSS libraries provide communication between code (syscalls) and system databases found in /etc/ like passwd, group, hosts, services.

getent is the frontend command for getting info from these databases – and more.

getent --help should show all supported databases.

So instead of doing

cat /etc/hosts | grep myhostname

use

getent hosts myhostname

This should work for domain names too, because the fallback for the hosts file is dns.

The file /etc/nsswitch.conf specifies the order in which the databases are searched.

So getting an IP for a host is as simple as running getent hosts cats.com

dig

dig comes with the dnsutils package and provides useful DNS debugging information. For example, it will give you the type of record, nameservers and much more.

See all the root servers

dig

Doing a regular lookup

dig cats.com

Getting the short answer

dig +short +noall cats.com

Getting nameservers

dig NS stackoverflow.com

Doing a reverse lookup

dig -x 74.125.196.102

Using another resolver

dig @8.8.8.8 reddit.com

Tracing a DNS lookup (this is really cool)

dig +trace youtube.com

Tracing a DNS lookup but the output is YAML

dig +trace +yaml yaml.net

nslookup

The functionality of nslookup overlaps dig in many ways, but they are both good tools for poking at DNS. nslookup is also on Windows by default.

A regular lookup (A Records)

nslookup yahoo.com

A reverse lookup

nslookup 98.138.219.231

Getting nameservers

nslookup -type=ns yahoo.com

Getting start of authority

nslookup -type=soa yahoo.com

whois

Finally, there is the famous whois tool, which is used when you want all the details on a domain. In addition to DNS information, it will give you all kinds of info about the registrar, domain creation date, contact info, cities, domain and registry IDs and anything else on file.

whois is different than the other DNS tools, because it uses a separate protocol designed for querying databases, and hence the reason it can give so much more information. The accuracy of this information is never guaranteed, but can still be useful.

History

whois goes back to the infant days of the internet. I’ll leave this link to the Wikipedia page for anyone curious about the history or underlying protocol.

Domain Privacy

When registering a domain, there is typically an option to have the registrar protect your information – specifically when somebody performs a whois on your domain.