The OSI Model: Still Relevant?

Nearly anyone studying networking will encounter the OSI Model and the TCP/IP Model early on. These are often presented as models, not absolutes, and there are many protocols that don't cleanly fit in the model, which becomes quickly apparent the more you look at it. In an earlier post, I discuss how formal standards often seem immutable, but over time it becomes clear that standards are created by humans and must be kept relevant to be meaningful. I also discuss how it's important to know why something was originally put in place prior to proposing an alternative.

Last year, I found The OSI Deprogrammer, which claims that the OSI model is not "wrong but useful," but is in fact incorrect entirely, as it has been adapted from terminology meant to describe communication from "terminals talking to mainframes and circuit-switching networks of telephone monopolies." These were connection-oriented networks that functioned much differently than today's networks. I am not sure how long the above link will work, as eventually the author may publish the book, but he is distributing this link on social media platforms currently. (There are some minor grammatical mistakes and repetitiveness, but this is not meant to be a final copy in its current state.)

The book discusses how the original model referred to a single network, whereas modern networking is made up of multiple independent networks. When the OSI model was originally created, it was prior to the need for autonomous system number assignments, prior to TCP/IP, and many other things we take for granted as the foundation of modern networking.

It also contends that the OSI model "has the wrong abstraction level," that it "combines the detailed with the general." I think this is an accurate description of how students feel when learning it, in addition to something people can feel as they get further along. Most people first learn about the devices and cables themselves and perhaps some of the protocols at a high level, and many of the devices and protocols don't align directly or have to be forced into the model.

Additionally, the OSI model is often described as abstract or conceptual, but originally referred to very specific mainframe-related activities, hence the "retconning" of the model into today's network landscape. The book emphasizes that the OSI model misses is the concept that the local network is independent of the larger Internet.

Computers did not always have terminals for input. The upper 3 layers of the OSI model are present to represent what happens when an interactive terminal used an application that was running on the mainframe.

The book proposes seeing real packets flowing in Wireshark as a viable way to get people to understand what is actually happening, and how realistically sized information gets transmitted (e.g. not all at once). Wireshark has certainly been helpful to me in this regard, and it almost makes more sense to show it first: "Here's how real packets move." Then discuss something like: "Here's a model of how information flows, and there are parts you can see in Wireshark. However, it was originally designed to describe communication on mainframes in a way that no longer occurs, so take it with a grain of salt in a real-world scenario." Perhaps also something along the lines of: "Here's what you do when you suspect a physical layer issue," etc.

There's also a bit about the history of computing, ARPANET, and more. The book is very opinionated (I understand why) and not designed to be read from start to end necessarily (though I did.) It's definitely worth a look if you want to learn the true origins of the OSI layers and be able to explain them in the proper context (i.e. how irrelevant many of them are to the actual process by which a packet traverses the Internet today.) Spend some time jumping into The OSI Deprogrammer and form your own opinion!