Microblog Memes

But without it we will never get the Restaurant at end of the internet.

Series of tubes.

How's that rhyme go again? Remember, remember, the Eternal September?

I think I found the source of confusion.

The OSI model describes networking in general, defining a model in which almost all networks can be categorized and compared. This is important as hundreds (if not thousands) of standards and methods exist for handling each separate layer - some publicized, but many hidden/propriety/unpublicized.

Meanwhile, the TCP/IP model describes only a very narrow subset of networks, though it just so happens to be the most used kind of network - The Internet - is part of this.

This means that if you are working with the internet, then TCP/IP will likely cover all your needs, but as soon as you move onto more specialized or simply uncommon network types TCP/IP will be close to useless.

We could take an example:

I'm setting up a LoRa network between some neighbors, measuring stations, etc. The network will not be connected directly to the internet, so there will be no possibility of data transfer between the LoRa network and the internet until you reach the OSI Presentation layer.

The LoRa network will need to be much more efficient than the general-purpose internet-networking, and since only a few machines will be connected we'll use only a single octet as device identifier (alternative to MAC/IP, lets call it SoMAC), also having to write our own discovery service.

Likewise, we will make a new custom transport layer based on TCP, but with only space for that single octet identifier, no flags, no IP, and no checksums (I like to live dangerous). Let's call that SoSTRIP.

At last, we'll need to write some kind of socket to make sending the information easy, for simplicity we'll use a Unnamed Pipe.

Now, how would we go about representing this in both models (OSI vs. TCP/IP)?

In OSI we'd have:

• Physical Layer: LoRa (duh)
• Data Link Layer: LoRa driver with SoMAC discovery
• Network Layer: SoMAC addressing
• Transport Layer: SoSTRIP
• Session Layer: Unnamed Pipe
• Presentation Layer: ASCII can be whatever
• Application Layer: Cat

In TCP/IP we'd have:

• Link Layer: LoRa + SoMAC
• Internet Layer: No IP/Not Applicable/SoMAC(?)
• Transport Layer: SoSTRIP, except it isn't compatible with either TCP nor IP.
• Application Layer: Unnamed Pipe + ASCII + Cat

Please note that this is purely for demonstration purposes, as it's absolutely unfair to compare TCP/IP and OSI in this way, since they both are designed for different purposes, with TCP/IP being more popular but narrow, and OSI being more general but overly complicated for most use cases.

Also, please feel free to correct me, since it's been a hot moment since I had about OSI and TCP/IP in uni.

==EDIT== Formatting

But TCP and IP are layers in the protocol (that are actually used.) some of the layers are then effectively bypassed as you go further up to application layer.

It's good to understand theory and even though you can really only partially map TCP\IP onto the model, it helps with memorising and clarifying the different parts of a packet.

Unofficially, I think it also helps introduce the mindset of what happens to a lot of Working Groups and Standards by Committee. They can have all the best intentions in the world, but sometimes the hack works better and\or is more popular and you just have to get used to it. See Also: JavaScript, UPnP, X500 addressing scheme, dot1q, NAT.

Once upon a time there were more protocols than just TCP/IP. You could have IPX/SPX, Baynes and others.

Hell. Even now with the introduction of Quic, it’s starting to change the use of TCP/IP, so having a general understanding of the layers in an abstracted manner let you map them to the different syntaxes that are out there. Kind of like programming languages.

What do you mean "Theoretical" and "Never been used"? Are you writing this by sending off radio waves purely with your mind? Am I the only one using a modem and computer? (/j, but it seems to me that you're asking "why a plane needs engines and wings, when it already has a payload")

TCP (and UDP) just describe how to assemble the data into packages which can be somewhat reliably reassembled on the other end.

While it does have an address stamped on top (IP), it doesn't know how to get anywhere by itself. That's where the bottom 3 OSI layers come in (the physical wires - or wireless spectra/wavelengths - the data is transmitted through, the specifications of how the embedded devices talk to each other over these wires, and how to discover other embedded or other devices on a network). I can very much assure you that the wires do exist and are indeed in use.

Contrary, the upper layers are more about keeping communication going once a connection has been established.

It was a good idea when it was super niche. And people were using MUDS and newsgroups.

As soon as marketing and capitalism figured out they could use it to control media content and sell it to the masses, we began down this road where we are now, with a few billionaire monopolies running everything.

It had two waves. The first was in the 90s and it was about access. Starting with dialup AOL, genie, prodigy and all those free CDs in the mail. Then the telecoms wanted in and that was that, for a while. Then in another ~10 years, since they couldn’t control access (as that was then universal) they had to control backend content, and now we have the billionaire monopolies in the services themselves, X, Facebook, etc.

It had a good run there for a while though.

Well, it's good entertainment. But maybe it wasn't. Don't know what my life would look like without it.