You might have seen, in the recent pandemic, advertisements for masks where “one size fits all.” You might see the same sort of thing for gloves, or maybe even flip-flops—but what you will rarely find, in the networking world, is anything that fits all.
It is, in fact, challenging to explain the plethora of protocols designed and deployed by the networking community to solve any possible problem. There are at least fifteen different ways to perform a simple task such as distributing IP addresses to hosts, including DHCP and SLAAC. RFC2322 even outlines a mechanism where IP addresses are distributed using a peg and pieces of paper. In this system, each user physically goes to the server, which might be a large tent or a “small table in the corner of the server room,” and retrieves a piece of paper with information about the network portion of the address assigned. Along with this piece of paper, however, the user also retrieves a peg, on which is written a subnet, or an individual host address. The user then walks back over to the relevant host (sneakernet!) and manually configures the correct address.
Given the state of IP addressing in many organizations, using the peg mechanism might actually be an improvement in their management plan. Even if they use avian carriers, as further described in RFC2322, to distribute the slips of paper and pegs used to assign addresses. The one thing this RFC fails to describe, however, is how to return addresses for future use—but this would not be a for nearly infinite address space of IPv6.
Another instance of this kind of phenomenon is the speed of the network. Most network operators chase the latest hardware to make their networks faster. After all, faster networks create happier users. But—in networking, one size never fits all.
To counter the rising costs of network connectivity at these ever-increasing speeds, RFC1216 proposes a paradigm shift to Ultra Low Speed (ULS) networks. Given faster networks always cost more, it should also be true that slower networks always cost less. If you make the network slow enough, it should really cost nothing at all. Slowing the network down to this level, then, can create a paradigm shift in networking economics. By combining enough users over a single low-speed link, each at low enough speeds, the network might just become less expensive but generate money. The authors of RFC1216 note some postal services might be persuaded to use such a network, as it would not materially impact their current delivery times.
A companion, RFC1217, written at about the same time, responds to the request for research into the feasibility and use of USLs. Use cases include low-speed underwater communications (because transmitting signals through water requires very low frequencies, which means much lower bit rates and transmission speeds—underwater networks of this type are naturally ULS). Some proposals are made in this companion RFC to reduce the speed of transmission on a ULS network even further, such as minimum likelihood decoding, which chooses the least likely meaning of a symbol, reducing goodput to the minimal possible, and thus causing as many retransmissions as possible.
So the next time you hear someone say “what is the industry best practice in this case,” you can gently remind them that best practices imply a common set of problems and solutions—and in the networking world, one size never fits all. It’s in writing, it’s in an RFC, so it must be true.
Russ White has more than twenty years' experience in designing, deploying, breaking, and troubleshooting large scale networks. Russ is currently a member of the Architecture Team at LinkedIn, where he works on next generation data center designs, complexity, and security. His most recent books are The Art of Network Architecture and Navigating Network Complexity.