The first network addresses

As I mentioned, every device on a network must be uniquely identifiable so that, at any given time, requests intended for a specific device can be delivered. Likewise, unique addressing means that any responses can be reliably returned to the originating device, no matter how many network nodes lie between the two. If someone has written a service that solves a problem you have, it's only useful to you if you can actually use that service. This means either knowing the address of the device hosting that service, or, at the very least, knowing who to ask for the address.

Thankfully, this is a problem that was solved long before even the earliest incarnations of the modern internet. I am, of course, referring to telecommunication networks, and their well-established system for addressing and address look-ups. With early telecom networks, engineers needed to solve problems for a large set of devices that needed to be uniquely addressed. Whatever system they came up with, though, would need to exhibit the following characteristics in order to remain viable in the long term:

• Usability: The system would be used by anyone who wanted to communicate over the telecom network, and so the system could not have been prohibitively complex.

• Scalability: The ultimate goal was to connect every home in the nation with a single, unified network. The solution for adding nodes to that network would need to grow with the population and the geographic region that it would ultimately support.
• Performance: If a telephone call took as long as the postal service to deliver messages back and forth, no one would use it. And, while that was never going to be the case, there would certainly be a limit to what customers would tolerate in terms of speed and reliability.

Thankfully, what they came up with was a sustainable solution that has scaled and functioned for decades.

The system that telecom engineers devised was that of phone numbers. By assigning 10-digit addresses to each phone on a telecom network, engineers guaranteed a network capable of uniquely addressing up to 9,999,999,999 devices. Add two digit country codes to that, and you've got a network that can theoretically support up to a trillion devices, or over 100 unique addresses for each human on the planet, with approximately another 240 billion to spare.

You might have noticed that I specified that the phone numbering system only theoretically supports up to a trillion devices. However, there are certain limitations of the addressing system of telecoms that make reaching the theoretical maximum difficult. As most of you will be well aware, the first three digits of a US telephone number are known as the area code. Those digits were originally determined by the specific geographic location in which the phone was located. This helped route numbers quickly, but means that the total number of possible devices supportable by telecom networks is limited by the distribution of those devices across geographic regions. Within an area code, there is only a theoretical maximum of 9,999,999 possible devices; barely more than the total population of New York.

I'm over-simplifying the solution here, but what this trade-off provided for telecom engineers was a simple mechanism for narrowing down the possible field of physical phones to which an address would resolve as quickly as was reasonably possible. Calls could be routed to a substantially restricted region by examining only the first three numbers. This provided an obvious performance benefit by applying semantic meaning to a syntactic standard. The telephone numbering system simply specifies that a physical phone is addressed by a 10-digit address. That is a syntax requirement. However, the geographic information conveyed by the first three digits of that address is a semantic standard. There is an underlying meaning baked into those first three numbers that conveys how the entire address should be handled.

The scalability of this numeric addressing system helps network devices direct traffic accurately. For a human user, though, an arbitrary series of seven to ten numbers can be difficult to remember, and are error-prone to use. Those who grew up in the time before smartphones and built-in contacts lists may remember the need to have a Rolodex, or contact book, to keep frequently needed, but difficult to remember, phone numbers organized and on-hand at all times. However, it was often the case that you'd need to call someone whose number you didn't have conveniently stored. This is where the phone book comes in. It served as a simple way of mapping easily-remembered unique identifiers for a person (specifically, a full name and street address) to their corresponding network-friendly address (their phone number).

All of these features, taken together, provided telecoms with the hallmarks of a successful network implementation: usability (through the simplicity of phone books), scalability (with the extensive range of valid addresses), and performance (with improved routing speeds achieved by embedding semantic meaning into the syntactic standards of the address). By now, though, you've likely correctly guessed that we won't be programming for telephone networks in C#. So, let's see how the design decisions made by telecom engineers translate to modern computer networks.