0

In section 7.4.4.1.2 of ISO/IEC 7498-1:1994(E) it begins with:

In connectionless-mode, the Transport Layer provides the following functions to support connectionless-mode transmission:

But many web pages (1060 google results) have this sentence:

"but only connection-oriented communication in the transport layer"

The OSI model supports both connectionless and connection oriented communication in the network layer, but only connection-oriented communication in the transport layer.

Please refer to any web page that has this sentence and to the ISO document.

Please don't mention the TCP/IP model, UDP or TCP protocols they are totally irrelevant. (despite that they are in the tags as Ron keeps adding these tags)

  • @RonMaupin I want to know the reason why so many web pages say the opposite. – Ahmed Elsawy Apr 2 '18 at 20:54
  • You will need to ask the authors for clarification, but OSI layers 1 to 3 are almost universally connectionless, and layer 4 (transport layer) can have connection-oriented or connectionless protocols. Both TCP and UDP are transport-layer protocols, and TCP is explicitly connection-oriented, while UDP is explicitly connectionless. – Ron Maupin Apr 2 '18 at 20:59
  • @RonMaupin The data link layer has a connection-mode as mentioned in the ISO document in contrary to what you are saying. You may be referring to the implementation of the OSI model (TCP/IP) but that is not what I am asking about. – Ahmed Elsawy Apr 2 '18 at 21:17
  • I'm talking about real-world. For example, ethernet, a data-link protocol, is connectionless. Both IPv4 and IPv6, network-layer protocols, are connectionless. TCP and UDP are transport-layer protocols, and one is connection-oriented, the other is connectionless, and that flies in the face of the quote that is the basis of your question: "...but only connection-oriented communication in the transport layer." – Ron Maupin Apr 2 '18 at 21:22
  • @RonMaupin Yes, but this happened by chance. I know that UDP is connection-less. But I was asking about the OSI model not the TCP/IP model. – Ahmed Elsawy Apr 2 '18 at 21:25
5

First, the OSI model is just a model, and things in the real world do not necessarily match. In particular, OSes, do not implement the Session and Presentation layers separately from the Application layer.

The network (Internet for the TCP/IP model) layer is connectionless. The two most-used transport protocols are TCP and UDP. TCP is connection-oriented, but UDP is connectionless.

Since the OSI model is an ISO standard, the ISO/IEC wording is what you should go by.


As you read in the ISO/IEC 7498-1:1994(E), the Transport Layer is defined as both connection-oriented and connectionless. There are explanations about using it both ways (I highlighted the relevant text in this section, but the entire document speaks to connectionless-mode):

7.4.3.1.1 The Transport Layer uniquely identifies each session entity by its transport-address. When providing the connectionless-mode service, the Transport Layer provides a connectionless-mode service which maps a request for transmission of a transport-service-data-unit onto a request to the connectionless-mode network-service. In connection-mode, the transport-service provides the means to establish, maintain, and release transport-connections. Transport-connections provide duplex transmission between a pair of session-entities (through transport-SAPs).

The ISO definition of the OSI model is the definition of the OSI model. Anybody else who writes anything that disagrees with the ISO definition is making up something that is not the OSI model. Just because it is published in a web page does not make it true. It is true only if it agrees with the ISO definition of the OSI model.


Edit:

I found a paper that lays out the update discussions to the original OSI model. It appears that in the original model, there was no connectionless mode, even in the Network Layer. In any case, this was all hashed out well before the Internet went commercial in 1995, and any web sites that claim the Transport Layer cannot use connectionless protocols are simply incorrect. The current OSI model, available since before the Internet went commercial, has always included the connectionless mode. It's like reading the reports by doctors from the 1940s and 1950s claiming that smoking is good for you. I know people who still quote those doctor recommendations. Networking moves orders of magnitude faster than medicine, and the original OSI model was out of date before the Internet went commercial; it has been obsolete for 24 years. and anyone who cares about the original version far out of touch with reality.

1

If I understand your question correctly, Yes OSI layer allows connectionless communication OSI model defines the layers in the network model. But it dooes not specify if the the layers should or should not be connectionless.

As you are specificially asking for Layer 4.

UDP is connectionless. Also, It defines the layers and not the protocol to be used in those layers. Theoretically you can design your own protocol that works at the transport layer that may or may not be connectionless.

1

The models are ways of providing a framework/definition of abstract concepts. The reality of implementations don't (and sometimes can't) follow the constraints of the models.

Does the OSI model support connectionless communication in the transport layer?

I think you answered your own question as you quoted the relevant section of the ISO/IEC 7498-1:1994(E) standard.

But many web pages (1060 google results) have this sentence:

"but only connection-oriented communication in the transport layer"

I believe the section you quoted from ISO/IEC 7498-1:1994(E) may have been modified from a previous version of the standard to adjust the standard to the reality of UDP. Unfortunately, I can't find the copy I used to have of ISO/IEC 7498:1984, or its two amendments/corrections, to confirm this (not willing to go through all the floppy disks I still have to find this).

This would explain the widespread appearance of this conflicting information; especially as many people often refer to the derivative works of others rather than going directly to the source themselves. This same effect has also sometimes lead to inaccuracies in some information being propagated.

However, since these are simply conceptual models, this is not an actual problem. Consider for a moment the TCP/IP model. This model does not have a "definitive" source for the model, but is generally considered to be based off RFC 1122. This has resulted in a situation where some people consider the TCP/IP model to have four layers and others to view it as five. Additionally, people have given the different layers different names that they believe it to be more representative of the nature of the layer.

None of these are wrong, but neither can anyone claim that their representation of the model is the correct one. Wikipedia provides a number of these on this table. Yet the TCP/IP model is still relevant and useful despite all these discrepancies. Although I personally believe this is a large part of why network professionals often gravitate toward the OSI model (which is defined by a standard).

  • I also think that it may have been modified. But I am looking for facts. – Ahmed Elsawy Apr 3 '18 at 3:47
  • 1
    @AhmedElsawy, it will likely be tough to get since the updated standard is nearly 25 years old and the ISO doesn't provide the option to obtain withdrawn standards. You would need to find a source with access to the 1984 version of the standard (along with the addendum and correction documents), and those are going to be very rare (far fewer professionals in any IT related field at the time makes the audience for such standards much smaller). – YLearn Apr 3 '18 at 3:55
  • To the point of the web sites that claim that the OSI Transport Layer is only connection-oriented, the Internet went public in 1995, a year after the latest edition of the OSI model, and those web sites probably did not exist for years after that. It could be a point of confusion for authors that didn't keep up for years, but the OP wanted proof for the connectionless-mode that is part of the current OSI model, when the proof should really be demanded of the web site authors, who proclaim something clearly contradicted of the web site assertions by the OSI model, itself. – Ron Maupin Apr 3 '18 at 3:55
  • @RonMaupin, true. However much of that information was available before the Internet went public (ah, the days of gopher...). Plus, the information on many sites was based on printed materials, which prolonged the use of statements like this even longer. – YLearn Apr 3 '18 at 4:00

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