1

In the "network" literature there are always explanations of this two reference models: OSI and TCP/IP.

The TCP/IP conceptual model is indeed a "fact" in terms of implementation. Anyone can find the implementation of several TCP/IP protocols such as Torvalds IPV4 implementation.

My question is... is a specific protocol "tied" to a reference model? i.e: Is Ethernet only a TCPIP link layer protocol?

Are there any OSI implementations? Are there any OSI specific-protocol implementations?

We always speak where to fit a protocol (the level of it inside the TCP/IP stack). But we can read about "level 7" firewalls (and level 7 in that sense means "application" firewall, inside an OSI model... not TCP/IP).

2

OSI is a theoretical model defined by ISO and ITU-T and published in 1984 that defines layers based on different functions of the communication process.

From 1984 to these days, a lot of technologies have been developed, some by standarization and some of them pushed by the market needs.

When telecommunication companies as Cisco, for example, develop some new technology, the competitors try to get the same technology and most of the times they converge to a common solution making a de-facto standard.

Because of that, in real life the protocols are aimed to resolve specific problems, and then, they loosely adhere to the OSI model.

TCP/IP was developed keeping in mind mainly layers 3 and 4 only, because the main concern was routing and session connectivity. In that moment nobody thought of problems like IP adresses scarcity, for example.

TCP/IP is independent of layers 1 and 2, so you can have TCP/IP over Ethernet, WiFi, Frame Relay, ATM, PPP, etc.

During the 90s coroprate LANS didn't use TCP/IP, because Novell Netware was the main force in corporate servers and its native protocol was IPX.

IPX was also loosely based in OSI, but mainly designed to solve the communication problems between Windows PCs and Novell servers.

Layer 7 firewalls inspect the contents of network packets to deny / accept specific applications (streaming, bittorrent, etc). They're called Layer 7 firewalls in reference to layer 7 of OSI, but they run over TCP/IP networksk, so it's a "wrong" denomination.

| improve this answer | |
2

The TCP/IP or DoD model is a somewhat simplified approach that centralizes on the network layer. Lower layers are just the "link layer" which IP doesn't care much about.

The OSI model is an attempt to structure everything into layers, from the cable to the application and transported data. While it is/was used in some areas, it's sometimes simply not practical to adhere to it completely.

In reality, the OSI model is a good approach to the lower levels but beyond layer 4 it often starts to become unpratical or somewhat blurry.

But even the OSI model doesn't catch all subtleties: the physical layer and the data link layer are often divided into several sublayers in order to structure development in several fields. E.g. in Ethernet, many physical layers for the same speed use the very same physical coding sublayer (PCS) and just differ in the physical medium attachment (PMA) and the physical medium dependent (PMD) sublayers.

Why do we go to these lengths you ask? Well, the beauty of it lies in the combinations that you can use right out of the box.

You can connect your smartphone through wifi, bridge (L2) that connection to Ethernet on the access point, route it (L3) onto DSL (L1 & L2) to your ISP, we-don't-really-know-what-happens-there, eventually the same data comes out on the other end of the world where it turns out to be a photo you've sent (L7+). We have no idea what media were used on the way (fiber, copper, radio) and don't even really care - and this is the neat trick.

| improve this answer | |

Not the answer you're looking for? Browse other questions tagged or ask your own question.