This week on the Gestalt IT Rundown, I talked about the plan by Let’s Encrypt to reuse some reserved IP address space. I’ve talked about this before and I said it was a bad idea then for a lot of reasons, mostly related to the fact that modern operating systems are coded not to allow 240/4 as a valid address space, for example. Yes, I realize that when the address space was codified back in the early days of the Internet that decisions were made to organize things and we “lost” a lot of addresses for experimental reasons. However, this is not the only time this has happened. Nor is it the largest example. For that, we need to talk about the device that you’re very likely reading this post on right now: your phone.
By the Numbers
We’re going to be referring to the North American Numbering Plan (NANP) in this post, so my non-US readers are going to want to click that link to understand how phone numbering works in the US. The NANP was devised back in the 1940s by AT&T as a way to assign numbers to the telephone exchanges so they were easy to contact. The first big decision that was made was to disallow a 1 or a 0 digit at the start of the prefix code. On a pulse dialing system that sends electrical pulses instead of the tones we associate with DTMF the error rate for having a 1 as the first digit was pretty high. It was generally ignored by the equipment. A 0 was used as the signal to a switchboard operator to get assistance. So the decision was made to restrict the 1 and the 0 for technical reasons. That’s the first technical debt choice that was made.
Next, the decision was made to make it easier for the CO switchboard operators to memorize numbers using mnemonics. That meant assigning letters to the numbers on the phone rotary dial (later keypad). Every number got three letters, leaving out Q and Z because they’re weird. Every number, that is, except 1 and 0. Because they weren’t used as the start of a prefix code they didn’t get letters assigned as noted above. Every telephone exchange was named based on the numerical prefix, which could have a 0 or a 1 in the first two digits. For example, the exchanges that started with 47x would be named using a letter from the 4 followed by a digit from the 7. Those words started with GR, such as “GRanite”, GReenwood”, and so on. it was handy to remember those as long as the telephone system didn’t get too big. If you ever watch an old movie from the 50s and hear someone asking to talk to a person whose number is KLondike5-6000 that’s why. We’ll come back to KLondike in a minute. The use of words instead of number was slowly phased out starting in 1958 because it was just easy for people to use nothing but numbers.
So what about dialing outside of your local exchange? Well, for that you need an area code right? But that area code looks just like a prefix code. How can you tell someone that you want to dial a different state? Well, AT&T came up with an interesting rule there. Remember how no prefix code started with a 0 or a 1? AT&T said that area codes MUST have a 0 or a 1 as the middle number. That organization allowed for the switchboard operator or signaling equipment to know that the first three digit code was an area code, followed by seven digits for the prefix and handset number. Simple, right? Combined with using the 1 digit to signal a long-distance interchange call you had a system that worked well for a number of years. Almost fifty, in fact, from 1947 all the way to 1995.
What happened in 1995? The telephone administrators realized they were going to run out of room given the explosion of phone numbers. The phone companies, now more than just AT&T, realized all the way back in the 1960s that growth would eventually lead to them needing to do away with the rules about restricting the middle digit of an area code to a 1 or a 0. That’s why area codes created after 1995 have middle digits in the 2-9 range. Thanks to things like mobile phones we doubled or tripled the amount of phone numbers we were going to need. Which meant tearing up all those careful plans about how to use the numbers that were created back in the 1940s to solve technical challenges.
To The Klondike
What about that KLondike number I mentioned earlier? Well, KL is 55 on the dial/keypad, so a KLondike5 number actually starts with 555. Most movie buffs will tell you that any number that has a 555 in the prefix code is a fake number since that’s what you hear in movies all the time. Originally these number were assigned to local exchanges only and used for testing purposes, except for 555-1212 which is a universal directory assistance number in every area code. In 1994, people realized that there were a huge amount of numbers that could be added to the NANP pool by reclaiming those test numbers.
However, if someone gave you a 555 number in a business setting or at a bar or club, how likely would you be to say that it was a fake or invalid number? Likely pretty high given the preponderance of usage in film. In fact, 555-0100 through 555-0199 were still set aside for “fake” number or entertainment purposes, not unlike 192.0.2.0/24 being reserved for documentation purposes.
The 555-XXXX number range is the perfect example of why a plan like Let’s Encrypt and their suggestion to reclaim the space is going to ultimately fail. You’re going to have to do a significant amount of programming to get every operating system to not immediately reject 240/4 as invalid or not immediately assume 127/8 is a loopback address. Because those decisions were made long ago, relatively speaking, and are going to take time and effort to undo. Remember that AT&T knew they were going to need to change area code rules back in the 1960s and it still took 30 years to put it into practice.
Moreover, the plans by Let’s Encrypt and others seeking to implement the Schoen draft in the IETF are ignoring a very simple truth. If we’re going to spend all this time and energy rewriting half the networking stacks in use today, why aren’t we using that energy to implement IPv6 support universally instead? It’s already a requirement if you interface with the US federal government. Why are we going to spend so much time fixing a problem we know is broken and not scalable in the future? Is it because we’re holding on to some idea from the past the IP addresses should be four octets? Is it because we want to exhaust every possible resource before being dragged into the IPv6 future? Or is it because after all these years we don’t want to admit that a better solution exists and we’re just ready to move to the next version of IP after IPv6 and we don’t want to say it out loud?
Technical debt is the result of decisions we made at the time to do what needed to be done. It’s not malicious or even petty. It’s usually what had to happen and now we have to live with it. Instead of looking at technical debt as a crushing weight we need to decide how to best fix it when necessary. Like the NANP changes we can modify things to make them work better. However, we also need to examine how much work we’re doing to perpetuate something that needs to be rewritten and why we choose do it the way we do. Phone numbers are something that are going to persist for along time. IP addresses aren’t. Let’s fix the problem the right way instead of the comfortable way.
Nice article. The el reg link though takes you to a different page.
Thanks! Fixed the link to go to the right story.