Talk:OSI model/Archive 1

Copies content
A search on Google shows that this page has paragraphs found multiple times accross the internet (right down to the space in 'a re' in the last paragraph). Whether this means its in the public domain, or that they are all stealing from the same source I am unsure. If its the former, something needs to be cited. - Eean —Preceding unsigned comment added by 24.12.199.xxx (talk • contribs) 04:18, 17 December 2001 (UTC)
 * Much of it seems to be from http://www.btconline.net/~derrellh/OTI/OTI_Classes/CIS140/OSI_model/OSI%20model.htm. It seems highly doubtful that this is public domain. I've removed the content for now. --Stephen Gilbert

OSI Graphic - Osi model com.jpg
I don't think this graphic is appropriate for an encylopedia. 1. The reference to "The Internet" is not appropriate. The internet is not always network (ie it can be GSM, non-internet TCP/IP nets etc) 2. The figures of two human faces do not add to the information content and can infact be misleading (ie endpoints are not always, human). 3. This images does not bring anything new or significantly aid in depicting the concept, especialls since "Rm-osi parallel.png" and 2 two tables included in the article.

Recommended Action: REMOVE Image "Osi_model_com.jpg" from the article & restore formatting.

Note to Markolinsky: It seems you have spend quite a bit of effert on this drawing and it looks good, but I don't think the graphic style and content is not appropriate in this article. In the future, it is probably best to discuss in the talk page before doing changes.

--202.161.20.46 08:42, 27 September 2006 (UTC)

Removed from article. As said above - Internet usually just works on network layer. --Nux (talk)

Other ideas
from user talk:MyRedDice

Here's my thinking:

First on OSI you'll notice the 7-long table with some examples of things at each of the 7 OSI levels.

You'll notice ethernet is on level 2 and 10BASE-T is on level one.

goto 10BASE-T and you'll notice that its said to be a "varient of ethernet",

PLEASE help me clarify how a variant of ethernet is on a DIFFERENT OSI level than it.

(see, i think that OSI categorization is just a loosely thrown together piece of ill-defined cruft) Anyway, maybe you can educate me

I think I can help. Ethernet is level two. But it depends upon level one: the physical layer. Thus you can run ethernet either on 10BASE-T cables, or on 10BASE2 cables, or on please respond to [[user_talk:hfastedge so that I dont have to constantly look at recent changes.

anyway, "you're just changing which variant of the physical layer it works on" that still doesnt address how "10BASE-T is a variant of ethernet" as mentioned in 10BASE-T.


 * Have replied there - see new version of 10BASE-T :) Martin
 * You almost got it right but you responded to user page instead of user_talk:hfastedge . ONLY when u respond to the talk page do you get a little * eg: Hfastedge (Talk*) allowing you to globally be aware of new talks. Finally, you've clearly introduced data within the article that is conflicting (within the confines of the article). I'm going to put the article on watch, and see if anyone interested finally (and solidly) gets it "right".

The confusion here is that Ethernet is both layer 1 and 2. 10BASE-T is just one implementation of layer 1 for Ethernet. It isn't a "variant", it's a specific implementation. --AMillar


 * Yes. I've added some stuff about that in the section on layer 1.  At layer 1, there are a number of physical layers of Ethernet, from the old 10BASE5 to shiny modern 10 Gb Ethernet.  At layer 2, they all use the same 14-byte MAC layer header, with optional 802.2 LLC header. Guy Harris 02:06, 31 January 2006 (UTC)

Problems with the OSI model
I'm thinking that maybe I should add some of my comments at the end of Talk:Internet_protocol_suite (about how the OSI model really divides things up in functional layers, and you shouldn't expect the actual dependency maps of real protocol suites to follow it slavishly) to this page; that will prevent a lot of confusion as to e.g. why BGP, which uses TCP, is at the network/internet layer.... Noel 12:40, 18 Sep 2004 (UTC)

BGP is an application (Layer7) that uses a TCP/IP session between two "speakers" The fact that it exchanges routing information is not relevant for Layer3. The given that those "speakers" live in devices that are called Routers, does not degrade BGP from Layer7 to Layer3. Similar to Telnet/SSH, FTP to a router. Would Telnet to a Switch suddenly make Telnet a Layer2 function? I see no problems with the OSI model, only human errors in interpretation.

The Management Annex to the OSI Reference Model defines two kinds of management, system and layer. System management is, indeed, an application, such as SNMP in the Internet and CMIP in OSI protocol stacks. Layer management protocols affect the functioning of a given layer without the need to have something at the application layer; this is also discussed in the OSI Routeing [sic], ISO/TR 9575. The key attribute of a layer management mechanism is that its payloads affect the information transfer protocols of the layer, not that it run "on" the layer. BGP runs over TCP, but is layer 3 management. Pure OSI IS-IS runs directly over layer 2, but is layer 3 management. RIP runs over UDP, but is layer 3 management. OSPF runs over IP, but is layer 3 management. Hcberkowitz 02:47, 10 May 2007 (UTC)

Layer 0
Sometimes a layer 0 is introduced to refer to the physical medium the data is transported accros, eg. the copper or glass cabling itself. This is because strictly speaking the cabling does not belong to the osi model itself. [JP]

"If not this, layer 8 is another way to reference user error. 166.70.62.200 (talk) 22:43, 23 March 2008 (UTC)"

MPLS
MPLS is considerd to be a switching techonology, i.e. layer 2. However it runs on other layer 2 technologies such as Ethernet or ATM, then why is it not considerd layer 3. Then again is what defines a layer 2 Protocol one that specifies the next hop in the path while layer 3 specifies the final destination.

Possibily could MPLS be considerd a sub-layer of Layer 2, so if MPLS ran over Ethernet there would be 3 sub-layers LLC layer, MAC layer and MPLS layer. -Vec 19 April 2005


 * Yet another perfect example of why the ISO model sucks for explaining anything above framing (e.g. HDLC) and below transports (such as TCP). In a model which had separate "internetwork" and "network" layers, MPLS would be the lowest sublayer of the "internetwork" layer (since it is intended to be able to forward traffic across multiple networks, which may be different kinds of networks). Noel (talk) 18:07, 14 September 2005 (UTC)


 * Furthermore this is why some call MPLS a "Layer 2.5" because it's both above layer 2 and below layer 3. Cburnett 04:28, 2 June 2006 (UTC)

MPLS is a Layer3 function, without doubt. Not layer2, nor layer2.5  It modifies layer3 headers. The fact that it does it in Layer2 hardware (the switch) does not mean it will work in ANY switch. The switch that can act on MPLS fields in packets is relying on the routing info from routers at the edge of the MPLS autonomous system.

The confusion here is first due to people trying to coerce protocols into the original seven-layer structure defined in ISO 7498, without any of the relevant appendices, or, more importantly, later ISO work such as ISO 8648, "Internal Organization of the Network Layer". That document comes up with a more flexible three-sublayer model with the protocol-independent network layer protocols like CLNP or IP on top, a subnetwork protocol (i.e., lower layer, not an IP subnet) such as LLC at the bottom, and mapping protocols like ARP in between.

MPLS, however, doesn't neatly fit either the classic OSI model, or even the classic IP stack. It uses routing protocols (i.e., layer 3 management) to find potential paths, end-to-end management protocols like RSVP-TE to map the MPLS topology, and then hop-by-hop protocols such as LDP to manage the tables of the label switched routers.

Things get even more confused when dealing with Generalized MPLS (GMPLS), which doesn't limit itself to paths for packets/frames, but also for lambdas/optical wavelengths, time slots (e.g., SONET/SDH or other multiplexing), or even physical ports on a cross-connect device like a DACS. Hcberkowitz 02:55, 10 May 2007 (UTC)

MPLS is a routing mechanism that overrules the original protocols. The forwarding of packets or frames is done based on the lable added to the header instead of using destination address based look-up tables. Therefore MPLS is a Layer 3 protocol that changes the forwarding/switching at intermediate nodes. GMPLS extends the label switching concept to include bit-stream and capacity switching (also known as layer 1 and layer 0 in the IP centric world). However, GMPLS is mainly seen as control plane issue intended to provide a unique control plan that is able to manage the switching options of all potentially involved (sub-)network layers. Sprawl 12:39, 24 May 2010 (UTC) —Preceding unsigned comment added by 85.127.103.152 (talk)

some of the mnemonics are wrong (grafitti)
I am not changing it though, cause I am starting a new job tomorrow and I have to read some suff to feel ready. &mdash;The preceding unsigned comment was added by 64.231.152.31 (talk &bull; contribs) 03:28, 3 August 2005.
 * These were long since removed, the new ones work. --67.183.217.186 09:49, 9 May 2006 (UTC)

methinks a few mnemonics are misogynistic
Guys, can we consider deleting references to dicks and pussies? It really detracts from an otherwise useful page. &mdash;The preceding unsigned comment was added by 216.184.31.184 (talk &bull; contribs) 23:26, 3 August 2005.
 * Wikipedia is not censored. However, these were long since removed, and the new ones were bland enough for a Dummies guide (and probably somewhat less memorable, if more verifiable). --67.183.217.186 09:49, 9 May 2006 (UTC)
 * How about just using a single line for the ones currently there? Crude ones may be bad, but ones that span nearly two screenfuls for a relatively simple concept detract from the core purpose of the article are nearly as bad. I'd go to the article on mnemonics if I wanted to see them repeated in full that many times over! Just have a simple bulleted list like this:
 * A; B; C; D; E; F; G
 * A; B; C; D; E; F; G

Making sense of the OSI model
As I studied the OSI model, I was quite confused, and just didn't get the idea right. Eg, I tried also to map OSI on other protocols, and it never seems to work out. As such, for a long time I saw OSI more or less as an academic toy. It took quite some time before I finally got the sense of it, something I wanna share with you all.

First of all I think the OSI concepts are much more useful then the implementation of the model itself. What are those concepts? Well there is communication, there is layers and there is language. These work in close harmony. The essence here is that each layer communicates with the same type of layer at the other site of the 'communication line'. Let me stretch this: two layers of the same type are talking to each other in the same 'language'. You can see this in almost every picture, but it is never really emphasised. What is the use of this concept? Well it simply helps you to focus on the function of a layer: what does it do, or when designing: what does it have to do.

Next concept is that in the OSI model, except for the physical layer, all layers are separated, and thus actually can not communicate with each other directly. From the model it is obvious that layers are talking too the next lower level (through the so called Service Access Point or SAP). The essence here is that the layer has some means to talk to his counterpart. In other words it tells you how a layer can talk to its counterpart.

Last concept is language. I use the concept of language in the sense of a set of words related to the same domain (for the intended usage database and road don't belong to the same language; road and car generally will). To define a language I ask myself 'what words do I need', 'is the language complete', 'is the language consistent', 'are there any ambiguities'? In every day sense it is sort of pragmatic, semi-formal tool. In the OSI model language is not explicitly defined as a concept, but it's left implicitly. I'm not going to give a precise definition, however you can see as a set of words. These words should be the only words you use to describe a function. And yes: in this way there are many, many 'languages'.

Summary: layers on the same level are virtually functionally communicating to each other in the same language ('horizontally') by really communicating technically ('vertically') to the next lower level layer. These two forms of communication are more or less independant, and to keep things simple you should always focus on one of them at a time.

To me the OSI model is a generic model which solves probably most or all network problems. For my day-to-day work, this is not very useful. So how do I use this concepts? Well, as simple and as obvious as they seem, I use them a lot. First I used them to understand network protocols. It especially helps me to keep out of the OSI-mapping discussion (try to map tcp/ip on these concepts ;-), and it helps me to explain network protocols to a lot of people with ease. Furthermore, if I have to do some work on communication I always ask myself: what layer(s) are we talking about, what language is used and how do they communicate with each other. And yes: this sounds simple, but in my experience 4 or 5 out of 10 don't grasp the idea and take much longer time then needed (if they succeed at all) to complete a job. And this is for network people.

And now the fun part. You can use the same concepts in application development: every time you're working with interfaces, you're talking about communication. The obvious usage is of course for interfaces between processes. Especially during design you can get much clearer discussions. If you 'know where you are' in a protocol stack you can much more easily focus then when several concepts are mingled. A nice example is the man/machine interface. From the above you can conclude that, functionally, only layers who 'speak the same language' can talk to each other. A second issue is that each application has its own 'language'. Working with a financial application is something completely different from working with a video-editor. Each application has its own set of functions, concepts etc. and the human to work with it has to speak the same language. From a design point of view I've turned this around: besides using use-cases etc, in a pragmatic way I formalise the language that the intended users use. Interactively I ask them what they need from an application and I teach them how to talk to a computer. Crucial here is that the language which is developed here is actually their own language, but semi-formalised, and as such it is a language which can be understood by both layers. &mdash;The preceding unsigned comment was added by 213.84.162.175 (talk &bull; contribs) 09:02, 21 August 2005.


 * While the "meta-model" (the use of layers, etc) is fine, the actual ISO 7-layer model is totally useless and confusing for describing everything below transports (such as TCP) and above framing (e.g. HDLC). Cramming X.25, MAC, ARP, IP, ICMP, etc all into one layer is worse than useless. However, the model is widely known, and for some reason (terminal brain damage?) people keep using it. If it had just one extra layer ("internetwork"), it would be a much better model for use with current real-world networks. Noel (talk) 18:07, 14 September 2005 (UTC)


 * Unfortunately, no one except confused educators uses the raw seven layers. Even within ISO 7498, there is an annex that describes layer management, parallel to user information transfer, to each layer. Routing protocols and ICMP, for example, are layer management for layer 3. There is also a System Management function in the application layer.

ISO recognized the network layer could not be monolithic, and split it into three sublayers, the lowest of which overlaps layer 2. The top sublayer is independent of the underlying technology, which is perfectly consistent with IP and CLNP. The bottom sublayer, which is called "subnetwork access" using "subnetwork" in a very different way than IP subnets, can be 802.2, X.25 (with some caveats), etc. The middle sublayer, subnetwork dependent convergence, maps between the top and bottom, with such protocols as ARP.

The bottom subnetwork layer can be applied recursively, which allows such things, admittedly now obsolete, as 802.2/802.3 LAN emulation over ATM. 802.2 is on top, 802.3 next lower, a mapping between 802.3 MAC addresses and ATM addresses, and then link-local ATM protocols.Hcberkowitz 22:19, 5 June 2007 (UTC)

Removing the mnemonic devices
Removing the mnemonic devices was a bad idea. I came here looking for these (which I'd bookmarked) and found them gone. I dont believe that just because one person finds them hard to remember that they aren't useful.


 * I agree59.144.250.72 14:53, 23 August 2006 (UTC)


 * Please see the other discussion of the mnemonics on this talk page. Whether they are "useful" is not the concern.  The encyclopedia is for facts.  If you want useful tricks for learning things, try reading a Wikibook? ~ Booya Bazooka 15:21, 23 August 2006 (UTC)


 * A note that the TCP/IP model page links to the mnemonic part of this article. In addition, I was looking forward to reading this. (I use wikipedia more than google now...) —The preceding unsigned comment was added by Mimithebrain (talk • contribs) 17:23, 16 January 2007 (UTC).
 * Seconded. Funny how people take this encyclopedia idea far too seriously. Brettr 06:45, 18 March 2007 (UTC)


 * I am with Booya Bazooka. The mnemonic is really distracting.  It looks childish/amateurish.  The mnemonic has since been replaced.  Bpringlemeir (talk) 17:52, 29 January 2008 (UTC)

X.25 and OSI
As far as I have been able to tell, not all of X.25 is used by the OSI suite (X.25 is a CCITT standard), although some of the protocols (HDLC?) do seem to have been used for OSI. Perhaps X.25 doesn't belong in the OSI column in the table of examples? StuartBrady

It doesn't fit neatly, especially given X.25 is actually a 3-layer stack. The packet level of X.25 defines the Connection-Oriented Network Protocol in OSI stacks. Its definition is a bit circular, as this was defined by ISO to be able to bring X.25 into the OSI suite. The X.25 packet level (not layer) has some functions that are more properly in transport and even application. Hcberkowitz 06:49, 17 May 2007 (UTC)


 * X.25 is not really a connection-oriented network protocol: X.25 may be used to provide the connection-oriented network service. See X.223 for details. Dgtsyb (talk) 02:54, 18 June 2008 (UTC)

Links to seven layers
The seven layers are described on their own pages, e.g. Application layer.

May I suggest someone adds links from the OSI model article to these pages? Maybe in the section "Description of layers". -- Felix Wiemann 13:59, 14 January 2006 (UTC)


 * The links are already there &mdash; the first link in each subsection links to the article for that layer. One of the links was broken, but I've fixed it now. StuartBrady 19:14, 14 January 2006 (UTC)

Layer 2.5
I would like to question the usefullness of discussing a "layer 2.5" in this article. It is not part of the OSI model, nor were any of the examples of a 'layer 2.5' protocol designed for the OSI protocol stack. If this *has* to be mentioned, then this should be mentioned after a definition of the formal layers, perhaps even under an 'Informal Layers' header. -- Ryanfantastic 11:01, 31 March 2006 (UTC)


 * Disclaimer: I added it to begin with. That said, I see no problem in its location (which is why I put it there).  The "usefulness" of layer 2.5 is that &mdash; seemingly like all communication stacks/models &mdash; it points out the need for something between layers 2 & 3.  I also don't see the need to reside it to an "others" category: it clearly states that it's not apart of the model and it flows just fine to me. Cburnett 22:45, 31 March 2006 (UTC)

It currently states that MPLS operates on packets (layer 2)... should that not be (layer 3), or frames (layer 2) Thedarxide 09:42, 22 May 2006 (UTC)

This function is generally covered by the Internal Organization of the Network Layer, although MPLS does not fit, was developed after real-world protocol developers (as opposed to pedants who don't keep up with network engineering) gave up on a seven layer model, and will never fit cleanly into OSI.Hcberkowitz 22:18, 5 June 2007 (UTC)

Mnemonics
I have deleted all the mnemonics because how many are being made up. WP:V says the burden of evidence (sources) is on the poster. See here for the mnemonics deleted. Cburnett 05:03, 6 April 2006 (UTC)
 * You're going to wait for an official ISO standard mnemonic? 88.105.122.117 09:37, 16 April 2006 (UTC)
 * The editors of CCNA For Dummies supported a few, those are now added and cited as an external link. --67.183.217.186 09:49, 9 May 2006 (UTC)
 * Excellent. Any source was exactly what I was looking for. And, no, I'm not waiting for an official mnemonic (it'll never come). Cburnett 00:00, 10 May 2006 (UTC)

I deleted the mnemonics again. Just because you have a citation for them doesn't make them valid for inclusion; they're just some things that a For Dummies author made up. If you want to help people remember the facts given in this article, I suggest writing a Wikibook on the subject. For the encyclopedia, please, just the facts. ~ Booya Bazooka 05:29, 9 August 2006 (UTC)

Um. If it's not referenced, but isn't stupid either, you can keep it. Seriously, there's going too far and then there's going too far. Wikipedia is NOT a bureaucracy, Wikipedia is an encyclopedia. --Kim Bruning (talk) 05:05, 12 January 2008 (UTC)

Capitalization
I noticed in this article the names of the layers begin with a capital letter, however in the 7 articles on the layers themselves they are in lowercase (most of the time). This needs to be rectified. —Preceding unsigned comment added by 141.157.54.202 (talk • contribs) 21:23, 19 April 2006 (UTC)

Security services on layer 6 (encryption).
I was doing a research project on cryptography and in a certain text, (Mitchell, Chris J, Users Guide to Cryptography and Standards, Artech House inc, 2005 [p 32-33]) the authors state the following

"No security services can be provided on in layer 5 or layer 6, although layer 6 may contain facilities to support the provision of services at layer 7."

I must say that I agree with the text, that security services such as encryption are not handled in the functions of presentation layer. It is my personal view that the 6th layer serves the function of interpreter rather then decipherer and as being analogous to the interpreter should not have a hand in the content of the message, or data to be conveyed. Therefore I believe that the functions of (en/de)-cyperment should be excluded from this layer as it is seen in the word encryption.

Magus 05:19, 20 April 2006 (UTC)

HTTP Layer 7 ???
My view of the layers here is different: TCP Layer 4, HTTP: Layer 5 - It provides a session going beyond one TCP connection (i.e. session cookies etc..) HTML: Layer 6 (presentation) it provides means for prensenting a web page, Layr 7; Appication: IE/streaming applications/IPC etc... —The preceding unsigned comment was added by 67.183.217.186 (talk • contribs).


 * I completely agree. HTTP is a session layer protocol; same goes for the FTP protocol.  I suspect the confusion here started long ago when things like FTP were listed at the application layer/layer 1.  In fact, this was actually somewhat valid: the FTP application (which on most *NIX boxes is called simply "ftp") could perhaps be considered layer one (it isn't a protocol, but the OSI model doesn't only describe protocols), but regardless the ftp protocol belongs at the session layer.

As another example, consider SMTP, which is also a session layer protocol. The article shows MIME as being at layer 6 (which I agree with). MIME encoded messages are transported over SMTP, but SMTP belongs below MIME in the model. But the article shows SMTP above MIME in layer 1/application layer. Iambk 19:53, 22 November 2006 (UTC)


 * The OSI Model is a protocol stack, I.E. is not a protocol. Another thing to consider: how would your stack look with HTTPS instead of HTTP? Ryanfantastic 18:18, 9 May 2006 (UTC)


 * First, TCP/IP model is not the same as OSI. The comparison is done to make a comparison: that's it.  There is no presentation layer nor a session layer as it applies to TCP/IP and its protocol.s


 * Secondly, HTML is not a protocol. It's a document/file.  Period.


 * Lastly, I suppose you could, however, make an argument about XML-RPC and SOAP being protocols and can find a place in the stack. Cburnett 23:47, 9 May 2006 (UTC)


 * (I was the one initially started this section) HTML is a document, that does not mean it cannot be a protocol… however this is not the issue here. For example all Web services which use XML over HTTP, this for sure have more layers than 7…

Also, the statement …”TCP/IP is not same as OSI”… TCP (over IP) is a protocol just as any other protocol, there is not reason to take it out of the OSI model. The OSI model talks about protocol layers (not necessarily 7), for sure there are many layers of protocols on top of TCP/IP which makes it 100% compatible to the OSI model. For example, in MMS: MM4 is a protocol, over SMTP, over TCP/IP – see 3GPP specs. I can give you a lot of protocols very real on top of the “allegedly layer 7” in this article.
 * Ranc 08:00, 17 May 2006 (UTC)


 * An HTML file is a file that contains data and a basic set of instructions on how to display the data (markup). You'd have to really bastardize the meaning of "protocol" to make a fixed-content file be considered a protocol.  HTML is no more of a protocol than a term paper: it contains data with markup instructions.  Is LaTeX a protocol?  If you answered yes then you really need to hit the copious amounts of texts on that very subject.


 * TCP/IP is not the same as OSI. In fact, the OSI model was created after the TCP/IP stack was created.  OSI is typically used to explain the TCP/IP suite but that does not mean TCP/IP is OSI or that there must be a one-to-one relationship between them.


 * What you are completely missing is that the Internet protocol suite only defines four layers. Everything above TCP/UDP is in the application layer.  There is absolutely nothing saying that the application layer cannot have multiple layers to it but they are all still in the application layer.


 * At the end of the day you have to realize that the OSI model is a model. Models change.  Models don't fit all scenarios.  Models aren't perfect.  Cburnett 04:26, 2 June 2006 (UTC)


 * HTTP(s) as layer 5 makes perfect sense with HTML as layer 6, presentation. The explosion of the Internet in the mid 1990s was triggered by the World Wide Web and the Browser. The WWW (HTTP, HTML) solved the layer 5/6 problems with elegance - allowing a very wide range of devices to present content without having to share a common complex document format.  "Present" is the key term here since much of HTML's initial focus was how things look on the page.  —Preceding unsigned comment added by 71.181.40.33 (talk) 21:25, 9 September 2009 (UTC)

The mnemonic "A Perfect Student Needs To Drink Port" does not identify the correct letters for network and transport layers cDima 21:13, 13 June 2006 (UTC)

RIP is a routing protocol
I just reverted someone moving RIP from layer 3 to layer 7.

It's a routing protocol. I suspect that the mover had a different RIP in mind.

There - just wanted to have it on the talk page too. --Alvestrand 20:53, 31 August 2006 (UTC)

RIP might appear as an ambigious case, but I think "RIP is a UDP-based protocol." from http://tools.ietf.org/html/rfc1058 proves that it has to be above layer 4. It has to do with layer 3, because the information send through RIP will eventually be used to reconfigure the IP routing table, which is used in layer 3. But as a protocol, it is clearly above layer 4. Matthias


 * Sorry, no. The IETF doesn't design protocols to fit neatly into OSI layers, but trying to coerce a routing protocol into a given upper layer due to its carrier mechanism is a common error. BGP, ISIS, OSPF, EIGRP, and RIP are all layer 3 management protocols, although they are carried, respectively, by TCP, data link, IP, IP, and UDP. From the IETF perspective, the payload defines layering, but layering is not an IETF priority. Hcberkowitz 06:46, 17 May 2007 (UTC)

I agree that routing protocols are layer 7. They contain data intended for use by the routers themselves and are never used to carry data from other end hosts to each other. One does not assign a RIP or OSPF address to a host. One assigns an IP address to a host. One never uses RIP or OSPF to transfer HTTP data. One uses UDP or TCP. RIP, OSPF, and even ICMP are application protocols that actually contain data to be used directly by the recipients. McNuttJ —Preceding comment was added at 17:55, 13 December 2007 (UTC)

Images
At least one of these images needs to be removed, since they're so similar. I would want to remove them both, though - do the pictures tell us anything that our text tables don't? ~ Booya Bazooka 15:46, 28 September 2006 (UTC)

I agree they both should be removed. See my comments re 2nd pic above Talk:OSI_model --202.161.20.46 08:11, 29 September 2006 (UTC)
 * I agree - both should be removed. Your comments about the second picture are spot on.  The other picture is better but I don't like the bidirectional arrows between the layers as I believe they are unnecessary and confusing.  If those arrows were removed or at least explained then I might feel otherwise.  --ElKevbo 15:47, 29 September 2006 (UTC)

I think the image should be removed. It is beyond the scope of illustrating the OSI Model. The OSI Model is a logical model. I appreciate the efforts of the author. I think a solution is 1) remove current image 2) discuss a future image on the discussion page 3) make new image from input. Joneboi 06:00, 19 November 2006 (UTC)

Freely-downloadable ISO 7498-1 OSI Reference Model document
Is it correct, that the link http://standards.iso.org/ittf/PubliclyAvailableStandards/s020269_ISO_IEC_7498-1_1994(E).zip is pointing to a Zip-File containing a PDF, that itself contains a scan of the ISO-Specs which are under Copyright Protection? Please have a look at that. --80.131.151.219 21:57, 24 October 2006 (UTC)


 * It is, indeed, true that the ISO are allowing some standards, including ISO 7498-1, to be downloaded for free from the page at http://standards.iso.org/ittf/PubliclyAvailableStandards/. That page says:


 * ISO Copyright for the freely available standards


 * The following standards are made freely available for standardization purposes. They are protected by copyright and therefore and unless otherwise specified, no part of these publications may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, microfilm, scanning, reproduction in whole or in part to another Internet site, without permission in writing from ISO. Requests should be addressed to the ISO Central Secretariat.


 * The documents you are about to download are a single-user, non-revisable Adobe Acrobat PDF file, to store on your personal computer. You may print out and retain one printed copy of the PDF file. This printed copy is fully protected by national and international copyright laws, and may not be photocopied or reproduced in any form. Under no circumstances may it be resold.


 * I don't know why the particular standards offered from that page were chosen. Guy Harris 19:05, 22 November 2006 (UTC)

Proposed merge with "Open Systems Interconnection"
I'd say "no" - Open Systems Interconnection discusses the overall OSI project, while this page discusses a specific item, the OSI networking model, that the Open Systems Interconnection page claims antedated the project. Guy Harris 19:18, 22 November 2006 (UTC)

Support for not-merge. As per Guy Harris. Visor 14:53, 30 November 2006 (UTC)

Description section, no layer 1?
In the description section, where the seven layers are explained, why is there no detail for Layer 1 (Physical)? Sapbuckets 16:27, 13 December 2006 (UTC)


 * My guess is that zealous edits got rid of it. Unless there's some discussion for or against its removal it should go back in. Luis F. Gonzalez 21:16, 15 December 2006 (UTC)

It was deleted by a vandal on 12 December 2006 and not reverted correctly. I put it back. --Rick Sidwell 05:41, 27 December 2006 (UTC)

Description section, switches listed on layer 2 & 3
Switches are listed as being part of layer 2 and layer 3. I am confused.

64.211.50.62 16:38, 26 December 2006 (UTC)


 * It is confusing, and the article needs to be clarified. So-called "layer 3 switches" merge layer 2 and 3 functions in a single device. Remember the OSI model is just a model; real-world protocols and devices are not obligated to follow it exactly (and generally don't)! --Rick Sidwell 06:02, 27 December 2006 (UTC)

Moved BGP from Layer 7 to Layer 5
I just had a discussion about this with a highly qualified individual on the subject of BGP. The protocol creates tcp sessions which puts it in 5. Although the RFC [] doesn't explicitly state the location in the OSI model, nowhere does it describe an attribute which would put it in any higher layer than 5. An application that administers BGP does not make BGP an application.

I have also attempted reverse arguments to qualify BGP for Layer 7, such as it's route table being a stored data stream (Layer 6), which I could not fully validate. BGP however is a protocol, not an application. Comments please.


 * BGP is a routing protocol that uses TCP as a transport mechanism. To have it at layer 5 makes no sense. For consistency with the other information on this page it should be at layer 3 with the other routing protocols. RIP uses UDP as its transport, however it is not placed at layer 5. OSPF uses IP as its mechanism, should it be at layer 4. The concept of a routing protocol is to build routing tables to allow layer 3 forwarding hence they are ancillary protocols of this —The preceding unsigned comment was added by 63.138.206.34 (talk) 05:09, 16 January 2007 (UTC).


 * The reason why seemingly all routing protocols are at layer 3 of the OSI model is because the very purpose of layer 3 includes "Path Selection". BGP is a layer 7 protocol due to its unique properties, and the fact that BGP relies on other routing protocols (IGP's) to function. The exception of RIP actually having a UDP header and port remains layer 3 because it does not rely on other routing protocols to function due to its (broad|multi)cast nature. These layer assignments are well documented and justified in various (very thick) books from cisco systems. Please contact me if you wish to discuss this further. Metaclassing 09:10, 3 February 2007 (UTC)


 * I participate in BGP standardization work in the IETF, and also was active in ISO architecture. Routing protocols of all types, regardless of the protocol that carry them, are layer 3. The confusion here may have something to do with assuming that all protocols are user information transfer and part of a seven layer stack. The Management Annex to the OSI Reference Model identifies layer management protocols as having payloads relevant to a specific layer, regardless of how they are transported. Would the pure OSI Intermediate System to Intermediate System (ISIS) be layer 3 because it carries layer 3 information, or because it runs directly over the data link? The IETF certaintly does not try to coerce protocols into OSI layers, and overemphasizing a layer orientation for a management or edge signaling method is a prescription for frustration. BGP is a layer 3 management protocol, not layer 5, 6, or 7. Howard C. Berkowitz 06:39, 17 May 2007 (UTC)

I argue that BGP is a layer 7 protocol used to manage a layer 3 protocol. It is not possible to run BGP without tcp or sessions.69.68.125.6 (talk) 04:01, 16 December 2007 (UTC)

RIP/OSPF
Having RIP and OSPF on the same layer makes no sense. RIP messages are encapsulated into UDP packets, while OSPF messages are directly in IP packets. Hence, if OSPF and IP share a layer, so should RIP and UDP. —The preceding unsigned comment was added by 84.193.228.136 (talk) 19:34, 16 January 2007 (UTC).


 * RIP, OSPF, ISIS, IGRP, EIGRP, and BGP are all considered, in the IETF, network layer management protocols. You might do better to picture a second management stack paralleling the layers that carry user information. The routing protocols are assigned to a layer based on their payload -- network layer addresses and routes -- not how the payloads are carried. According to the Annex 4 (Management) of the OSI Reference Model, routing protocols are layer 3 management protocols. There is further elaboration of this point in the OSI Routeing [sic] Framework technical report from ISO. Howard C. Berkowitz 06:42, 17 May 2007 (UTC)


 * Those protocols are indeed network management protocols, but that does not make them network protocols. Network management is an application, not a network in itself.  Again, one assigns a host an IP address, not a RIP address or an ICMP address.  RIP, OSPF, and the like are applications that depend upon an active IP layer in order for two routers (hosts) to communicate with one another.  It would be well to remember that routers are merely hosts that are attached to two or more networks AND configured to pass data from one network to another.  The protocols they use to talk to one another do not create networks.  The simply allow the two hosts to trade information about other networks.  mcnuttj

802.11 listed multiple times in Examples section
In the Examples table, under "Misc. examples" 802.11(WiFi) is included as a layer 2 example while 802.11b & 802.11g are listed as layer 1 examples. This is confusing and should be clarified. I believe 802.11(WiFi) should be listed under layer 1 only.

TweakerTV 20:12, 30 January 2007 (UTC)


 * The 802.11 standard defines several physical layers as well as the MAC layer. So its both layer 1 and layer 2. --141.20.23.63 13:23, 11 April 2007 (UTC)

SIP Layer 7?
Inconsistent with SIP as mentioned in Layer 5. In the 4-layer/5-layer model, I suppose anything above the Transport layer is sometimes called an "application". But in the 7-layer model, seems to me that SIP is certainly a session-layer protocol; I understand it can handle arbitrary data structures at the presentation layer, and has nothing whatever to say about how applications handle those what's presented. For that matter, similar issue about http discussed above was not resolved. html is obviously a data structure at the presentation layer, no? If so, how could http be at a higher layer than html?

I can't decode the note in the table entry "TCP/IP Suite" for Layer 5: ''...SIP. (Not a separate layer with standardized API.)'' Not sure if this specifically refers to SIP. Whetever it refers to, I say that real world layers aren't perfectly independent of other layers as in the ideal model. OSI is very useful despite the fact that it does not map perfectly to the way life works. So I'm not sure what special non-separateness this refers to in terms of SIP or anything else. And I'm not sure what what point the editor was making about APIs.

I think there are nested stacks in the world: when tunnelling, inside Sonet...the possibility of nested stacks does not mean that outer stacks are necessarily at higher layers than inner stacks, and it does not necessarily mean that layers are blurred. YMMV. OT, I think SCCP is at SS7 Layer 4, not 3. --Michael Gold 00:49, 3 February 2007 (UTC)


 * There is a decent image from intel that compares the OSI model with the SS7 protocol stack. Metaclassing 09:14, 3 February 2007 (UTC)


 * ...which doesn't look right to me - while the TCAP is located at Application layer, it is rather Session protocol - all it does is structuring dialogue between upper layer protocols, it doesn't provide anything more than that. Also compare with this talk. Yacoob 16:32, 10 June 2007 (UTC)


 * ...I can't help observe that this is another case of trying to force protocols, which were not rigidly defined to follow the OSI model, into the OSI model. Among the OSI protocols themselves, the Remote Operations Service Element (ROSE) has a function similar to Internet Remote Procedure Call (RPC). Ironically, in a stack for the Internet Network File System, there are truly seven layers, with RPC at Session, XDR at Presentation, and NFS proper at Application.  AFAIK, there was never a full OSI implementation of NFS functionality, but rather the more complex FTAM. The Distributed Transaction Processing stack (OSI) is closest to SS7 TCAP.Hcberkowitz 16:50, 10 June 2007 (UTC)

European Domination
The history section claims that ISO is european dominated. With 157 country members, how can it be dominated by one contintent? I wonder if someone is bitter because 'their' country/continent doesn't dominate ISO? Markb 13:04, 5 February 2007 (UTC)


 * It's more fair to say that especially CCITT/ITU, which did some of the early work, was dominated by national telephone monopolies. In the seventies, that was more true of Europe and Asia. Howard C. Berkowitz 04:02, 21 May 2007 (UTC)

ishfaq kallill
What does this heading, between layer 3 and 4 mean? —The preceding unsigned comment was added by 62.145.29.194 (talk) 09:16, 16 February 2007 (UTC).

802.11
Just a note here to discuss my changes. 802.11, 802.11a, 802.11b, 802.11g, and 802.11n each have their own PHY, MAC and LLC entities. I.e. 802.11a/b/g/n are not only PHY layer specs, but include MAC/LLC also. And, the 802.11-97 (legacy) standard also included MAC/LLC and actually three different PHY layers: FHSS, DSSS, and Infra-red.Short description of the standardKgrr 23:33, 6 April 2007 (UTC)

Graphical analogy
There used to be a lovely on this matter on this page. This was removed. Anyone know why? It was really useful for non-engineers to follow. —Preceding unsigned comment added by 64.194.250.99 (talk • contribs) 22:06, 18 June 2007 (UTC)

Internal Organization of the Network Layer, and the Futility of Forcing non-OSI Protocols
OSI architectural development did not stop with the basic OSI reference model, ISO 7498. The original work at all layers, especially Network, was strictly connection-oriented. That reflects the circuit-switched predisposition of the telephone engineers that originally developed the model, and their realization of packet switching as X.25 virtual circuit switching.

The first addendum to the OSI Reference Model proper, 7498/1, introduced connectionless communications, as in the Internet Protocol. A later document, ISO 8648, "Information processing systems - Open Systems Interconnection - Internal organization of the Network Layer (IONL)", came in 1988. The IONL document attempted to make real-world protocols fit the OSI network layer, and recognized the reality that the strict modeling of Network and Data Link didn't work. I'll introduce a simplified version. Since it costs USD $102 to download the IONL document, you'll have to trust my memory.

''sorry, I don't know how to edit a Wiki table. Feel free to fix the attempt below''

Abstract name................Connection-oriented stack..........Connectionless stack

Subnetwork[note 1]...........X.25 packet layer, more or less....IP or ISO CLNP Independent

Subnetwork-dependent.........None (in reality, static mapping...ARP (for 802.2 LANs) Convergence..................between NSAP address and X.121

Subnetwork access............LAP-B..............................IEEE 802.2

[Note 1: In OSI-speak, "subnetwork" is a specific technology at data link and below. It is not the same as an IP subnet]

To make things worse, real-world protocols can recursively apply these sublayers. ATM LAN Emulation, for example, at the emulated level did an ARP over LLC, but then used ATM protocols (mercifully forgotten) to map the MAC address to an ATM address.

Routing protocols, incidentally, are in a parallel stack for layer management at the network layer. See the OSI Management Framework, ISO 7498/4, and the OSI Routeing [sic] Framework, ISO/TR 9575.

People really need to recognize that IETF and IEEE protocols were not designed to fit the OSI model, as much as basic networking instructors try to force, coerce, or shoehorn them into a 7-layer oversimplification.

To give a little personal context, I was a member of the Federal Telecommunications Standards Committee in the late seventies, when the ANSI Distributed Systems (DISY) project became OSI. Later, I was the first technical staff member of the Corporation for Open Systems (COS), the industry group for promoting and testing OSI and ISDN protocols. COS knew OSI was the answer, but we were still trying to figure out the question when IP protocols, for many reasons, passed us by. I was directly involved in OSI architectural discussions including liaison with IEEE 802, and later was the team leader for developing the CONS/CLNS/X.25 conformance test system.

Hcberkowitz 03:40, 20 June 2007 (UTC)

Impossible Clause
Read this from the introduction:

Even though newer IETF and IEEE protocols, and indeed OSI protocol work subsequent to the publication of the original architectural standards that have largely superseded it

Even though what verb? Does the writer mean "Even though there are newer...", or maybe "standards have largely superseded...".

I would change it myself but I don't get what is meant. Thanks

--Sukkoth 05:41, 20 September 2007 (UTC)


 * Good catch of an error. What do you think of "Without supplemental ISO work, or other protocol architecture work in the IETF and IEEE, the ISO 7498 model is obsolete for modern protocol development. It remains useful as an introduction, but real protocol implementations rarely follow it closely. The IETF, in fact, consciously avoids strict layering in the protocols passing the reality test of being the basis of the Internet. Howard C. Berkowitz 13:26, 20 September 2007 (UTC)


 * Well thanks for acknowledging. I wouldn't presume to evaluate the correct language here, cause I don't know anything about the topic--that's why I was reading the article. Your quote is at least clear and understandable. --Sukkoth 21:00, 27 September 2007 (UTC) —Preceding unsigned comment added by Sukkoth Qulmos (talk • contribs)

Layer 3 example
My removal of the metaphor example in the Layer 3 section was contested, and I'd like to discuss it here. While the latest rewrite has dispensed with any informality, I still don't think it's optimal. Specifically, the whole second paragraph in the network layer section discusses the Internet Protocol rather than the network layer. I think that the information contained within that paragraph belongs in the Internet Protocol article itself. I also think that an example in that section should explicitly discuss only Layer 3 of the OSI model. -FrankTobia (talk) 21:41, 9 December 2007 (UTC)

Fully agree and must add that the figures in this article do not match with the pure OSI model (ITU-T X.200). Sadly the IP community spoiled the elegant OSI model, and today if the OSI model is referred actually a mixture of IP and the original OSI model is commonly meant. This stub of a model has no meaning as it lacks a clear structure; it is simply a way to hierarchically classify protocols. The OSI model per se fits to Ethernet, SDH, ATM, IP, etc. equally; all these offer a Layer 3 responsible to offer an end-to-end connection. That IP lacks it's own lower layers demands to use other network layers as so called sub-networks to create an IP network. See the ITU-T X.200 reference (page 41ff) on the concept of sub-networks and how these integrate nicely to provide a clear view.

The Bit, Frame, Packet etc. termini must be cleared from the figures - these are not in-line with the OSI model, these refer to specific implementations. It is evident that every Ethernet LAN has all Layers, accordingly Ethernet frames travel end-to-end from application to application. —Preceding unsigned comment added by 85.127.103.152 (talk) 11:10, 24 May 2010 (UTC)

Router-to-router protocols, as well as router-to-host
Please do not put these in the application layer, where they do not belong. Let me give the relevant background, with the caveat that IETF/IP protocols do not follow strict OSI layering.

The original OSI Reference Model, ISO document 7498, did not address management at all. Annex 7498/4, however, introduced system management (e.g., CMIP and SNMP), as well as layer management. {| class="wikitable" ! Layer ! User information protocol ! Layer management protocol
 * Network
 * IP, CLNP
 * OSPF, IS-IS, RIP, BGP, PIM, DVMRP, etc., are all layer management between routers. ICMP, IGMP, and ARP are host-to-host or host-to-router, but still layer management
 * }
 * }

Other ISO documents further elaborated. The Internal Organization of the Network Layer document, ISO 8648, defined three sublayers, not counting recursive tunneling such as ATM LAN Emulation:

The OSI Routeing Framework, ISO/TR 9575, further clarified (using a European spelling for "routing") that router-to-router protocols are layer management protocols in the network layer.

For the record, I was the team leader for X.25 and ISIS protocol conformance testing at the Corporation for Open Systems, and subsequently work in the IETF OSPF and BGP (Inter-domain routing) working groups. I was the lead author of RFC 4098 on BGP control plane convergence.

If you want to move routing protocols into the application layer, please give authoritative sources that override the references I have just given.Howard C. Berkowitz (talk) 18:01, 13 December 2007 (UTC)


 * You can cite all the documents and credentials all you want, but you yourself have pointed out that the IP suite of protocols does not, and was never intended to, strictly match the OSI layers.


 * I believe Wikipedia requires the use of reliable sources. The OSI Routeing Framework, Internal Organization of the Network Layer, and OSI Management Framework all identify the protocols in question as layer management protocols of the network layer.


 * Therefore, we must fall back on the descriptions of the layers and how the protocols are implemented.


 * All right. RIP, OSPF, BGP were all defined by the IETF. Please produce one single IETF document that says the OSI layers were even considered in defining them.


 * Further, IS-IS runs directly over the data link layer, not IP or CLNP. Does that make it a network layer protocol?


 * RIP, OSPF, BGP, ICMP, and so on all use IP as their transport method. Some of them go further and use UDP.  None of these protocols create new addressing schemes or create new ways for two hosts to communicate.  They are applications running on two different hosts - often routers - that need to trade information.


 * Wikipedia expects authoritative sourcing. This is your original research, which is not supported by any IETF or ISO/OSI document. If it is not your research, produce credible sources to the contrary.


 * Furthermore, routing protocols are not a necessary part of a network. One can perfectly successfully run an IP network using static routes.  (Again, ARP breaks the rules because TECHNICALLY IP could be used without it, but IP is arguably useless on an 802 MAC network without ARP.)


 * IP is useless without ARP? You've said that routing protocols aren't necessary for IP networking because static routes could be used. I don't know of a major router implementation that does not allow static definition of IP to MAC addresses. By your logic, then, it's totally practical to run without ARP, because the mappings can be done with static definitions. Let me know when you've written out the static routes for an Internet inter-ISP router, and how well it works. What is its convergence time when a route changes in Tokyo and you are in Amsterdam.


 * This is not a complicated matter that needs endless citations. Their place in the OSI model - with the exception of ARP - is reasonably self-evident.


 * It may be self-evident to someone who has not been involved in OSI or IP routing protocol development, and apparently has not read the OSI documents that call these layer management protocols. Again, Wikipedia doesn't accept original research.


 * OSPF and RIP are routing applications that help routers make decisions. There are no RIP addresses or ICMP addresses.  No program uses RIP as a transport method to send data across the network.  The RIP application uses UDP to send data to another RIP router.  It's as simple as that.


 * Simple as that. What do OSPF, EIGRP, and ISIS use as transports?


 * I noticed you didn't say there are no OSPF addresses. What are 224.0.0.5 and 224.0.0.6? Howard C. Berkowitz (talk) 04:22, 16 December 2007 (UTC)


 * Wikipedia does ask you sign comments. —Preceding unsigned comment added by Mcnuttj (talk • contribs) 18:20, 13 December 2007 (UTC)


 * 224.0.0.5 and 224.0.0.6 are ipv4 addresses that are reserved as multicast ip addresses for use in OSPF. However, this does not make them OSPF addresses. Potentially, these ip addresses could be used for another purpose and other ip addresses could be used for OSPF multicast comminication. Examples of the latter can be seen when ospf is run over a point to point connection using unicast ip addresses and when OSPF is used with ipv6 where different ip addresses are used for multicast. Also, OSPF uses ip protocol 89 for transport rather than raw ip packets.69.68.125.6 (talk) 02:22, 19 December 2007 (UTC)


 * Yes, they are IPv4. If you check with IANA either in the data base, they were reserved at the request of the OSPF working group.


 * What is your point about it using IP protocol 89? Protocol 89 is not a transport protocol. It is a network layer management protocol. It seems very important to you to force IETF protocols into the OSI reference model, a model that was rejected by the IETF. Why is it so important? The OSPF specification writers and implementers don't feel it's necessary. The IAB and IESG don't find it necessary. OSPF's particular retransmission algorithm, on a broadcast multiaccess medium, was developed for routing protocols and has had some very limited use internal to fault-tolerant routers. Please find an authoritative reference that says OSPF uses an IETF-recognized end-to-end protocol for transport. Howard C. Berkowitz (talk) 16:12, 7 February 2008 (UTC)

Better Mnemonic
It seems sort of out of place on a wikipedia article but if it is there why not have one that goes in the order of the layers. We were told to remember All = Application

People = Presentation

Seem = Session

To = Transport

Need = Network

Data = Data Link

Proccessing = Physical

which goes in the protocol numbers from 7 to 1 which is better than the sausage pizza one which isn't really ordered TheGreatZorko (talk) 09:19, 31 January 2008 (UTC)
 * How is "Please Do Not Throw Sausage Pizza Away" unordered? It lists the levels from 1-7. Turlo Lomon (talk) 19:20, 28 April 2009 (UTC)

Small change DHCP
changed dhcp from network to application. It is where it should be. —Preceding unsigned comment added by Hellomarius (talk • contribs) 00:25, 12 February 2008 (UTC)

Comparison of Transport Layer to Post Office needs improvement/deletion
I'm talking about the comparison of the Transport layer to the Post Office.

First off, it doesn't really explain anything. Before the author actually makes his point, he then begins to ramble on about tunneling and GRE. There is probably a better way to visually explain this layer, but personally I don't know that this layer needs any visual explanations made at all.

Secondly, a "Post Office" comparison would be better suited for the Network layer. :-) Subfrowns (talk) 20:31, 22 February 2008 (UTC)

False History
References show that the SPARC/DISY work in ANSI was redirected to work on the newly formed ISO Open System Interconnect, not that the OSI Model work "started" in ANSI SPARC/DISY, even if personal papers (box of files) were a verifiable source, which they are not.

The term "OSI" could hardly have come into use on 12 October 1979 (that just the date on the box of paper, I believe), when ISO created TC97 SC16 charged with developing an "Open Systems Interconnect" architecture framework in 1977, and the "OSI model" was ratified by ISO in July 1979.

I have removed these two items which could not possibly be correct. Dgtsyb (talk) 04:41, 19 June 2008 (UTC)

I removed all reference to this person per WP:BLP. If someone has a problem with that, please let me known and I will file it on WP:BLPN. Dgtsyb (talk) 05:58, 19 June 2008 (UTC)

Colors
The colors in the OSI model image (Image:Osi-model-jb.png) on this page could need a change, since they are bad from the perspective of someone with any common color deficiency. Especially colors in the "Transport" and "Network" fields are hard to tell apart. Would fix it myself if it wasn't that I have both a green and red deficiency and would probably just mess it up. There's a chart with safe colors here and some more info here. -Manwal 23 June 2005

Spam Links
I think that the spam links are gone, so I removed the cleanup spam tags. Dgtsyb (talk) 04:46, 19 June 2008 (UTC)

Vote: Four and/or five layers in the TCP/IP model template and wiki articles?
Give your vote here. Should the TCP/IP model template have four or five layers? And what is the name of the bottom layer in case of four layers? And is it okay to mention both the four and five layer models in wikipedia articles? Mange01 (talk) 18:17, 17 July 2008 (UTC)

ASN.1
Can I suggest add ASN.1 as Presentation in SS7 stack? Asn1tlv (talk) 07:54, 4 September 2009 (UTC)

undo
OK, someone removed the additions of several more examples of stacks. Even when they have been placed in a separate section.

No talk, no improvements, just removal. This is not the way.

The current page contains no examples of OSI5, only one of OSI6 (ASN.1 with a not very good description). Interestings comments in this discussion about HTML/HTTP, SIP, ... have been ignored.

Nowadays that OSI3 and OSI4 are dominated by TCP/IP, most part of the improvements in network protocols are in OSI1-2 and in OSI5-6. But this page ignores them.

Lots of people doesn't understand the objective and usage of OSI5 and OSI6 layers. Moreover, they are not able of made a diferentiation between OSI7 and the final application. In conclusion, they decide OSI is not important, something of the pass.

It is clear this page will not help them. —Preceding unsigned comment added by 81.38.181.95 (talk) 16:06, 4 September 2009 (UTC)


 * The addition of that section was not motivated in discussion in the article itself and didn't discuss how this was different or 'alternate' from the existing discussion. It only added to the confusion that the Examples section already established. If your section meant to discuss individual application protocol layering from the ground up based on the OSI than this should be explained, but it is questionable whether that might be more appropriate for the individual, specific articles. Some of the categorization of protocols and alternate interpretations and the reasoning involved, may also be more appropriate in the discussion of each layer, as examples of layer implementation. Just presenting another complex table to show slight variations without discussion is not helpful at all. The existing 'examples' table already presents a lot of technical expert information without discussion. Kbrose (talk) 16:48, 4 September 2009 (UTC)
 * Really do you think that an example table of stacks where ASN.1 doesn't appears is complete?. Nor XML, ... . Where session doesn't contains XDR, SOAP, ... (but contains SDP!). Where HTTP is declared OSI7 without discussion, even when there are tenths of protocols who uses it as session one (see discussion chapter in this page)? Where someone can understand that ISUP uses SCCP as OSI3 or that SCCP is an alternative to MTP3? Where TCAP appears as OSI7 without discussion, even when their PDUs are "begin", "continue", "end" and "abort". No stack example with RTP,RTCP, ... nor any media/signaling protocols coming from new generation networks.
 * Anybody who reads it will decide that really the current networks has only 4 layers. That "session" and "presentation" are theoretical inventions discarded years ago and "application" is the executable file. For these reasons, I thought a second example table could be useful. Trying to not conflict with the existing, and accepting (can be) a classification less theoretical but more oriented to what be see nowadays. —Preceding unsigned comment added by Asn1tlv (talk • contribs) 17:15, 4 September 2009 (UTC)
 * It's obvious the existing tables aren't complete, nor can they ever be in a general purpose article, but anyone is welcomed to make improvements. The tables don't really provide discussion or rationale, they are just a summary overview and additional ones don't improve the situation. HTTP only occurs in the existing table under IP suite, where it surely is an application protocol, the mapping here could be improved by listing it in multiple layers. HTTP can surely be a lot of things not originally intended, but the details couldn't easily be conveyed in a table like this. I would suggest writing a real prose section of discussion somewhere, most likely the HTTP article, ditto for any other protocol that isn't treated fairly here. Kbrose (talk) 18:03, 4 September 2009 (UTC)
 * "[...]anyone is welcomed to make improvements[...]" It seems no, not anyone is welcomed. Because a second table, adding more examples without conflict with the existing one, has been immediately deleted. First, another col.collaborator had made a great job and corrected and improved it, but finally someone has deleted the initial work and their corrections. —Preceding unsigned comment added by Asn1tlv (talk • contribs) 12:09, 5 September 2009 (UTC)

Suggested addition
Although the OSI was never implemented in its protocals, it standardization efforts in defining "layers" was accetped. This article really misses the entire point.. TCP\IP only encompasses layers 4 and 3 of the OSI, protocal wise. And its perscription for the other layers lines up exactly with the that of the OSI. The OSI is much broader and its standard model is the basis for almsot all internetworking in the world today. You cannot take a networking class without learning the OSI model, it is the foundation upon which almost all internetworking is built.. The article seemingly dismisses the OSI as a failure, when in fact it one of the most critical and important things to understand in networking. &mdash;Charles Edward (Talk 18:39, 29 September 2009 (UTC)
 * (a) OSI was indeed implemented in actual protocols. (b) OSI is not so fundamental at all and was completely useless for internetworking as the first few independent implementations couldn't talk to each other. (c) TCP/IP development has been completely outside of OSI model. (d) none of the layers in the two models really line up exactly, because they are different models, the layering concepts are different. (e) it only appears critical for passing classes. (f) Internet architects have explicitly stated that OSI design is not a goal or foundation for any Internet standard. So what addition are you suggesting? Kbrose (talk) 01:26, 30 September 2009 (UTC)
 * Pardon my mistake. I intended this comment to be on much shorter and underveloped article, Open Systems Interconnection. Please disregard it. &mdash;Charles Edward (Talk 12:20, 30 September 2009 (UTC)

Moved L2TP to OSI Layer 2 from Layer 4
L2TP belongs to OSI Layer 2. L2TP just provides a tunnel that is then, for example, used by IPSec for for implementing a secure transfer of the packets: L2TP packet is wrapped and hidden within the IPsec packet. Also, by googling for L2TP OSI Layer it seems that Wikipedia is the only place where this protocol is placed at Layer 4. Also removed "L2TP carries PPP frames inside transport packet." from the description of Layer 4. —Preceding unsigned comment added by 195.145.148.246 (talk) 13:43, 5 January 2010 (UTC)

Most VPN technologies don't fit the OSI model well. VPN protocols provide a virtual layer 2 or 3 tunnel, but to do this they require the upper layers. L2TP provides a virtual layer 2 connection, but it requires layers 3 and 4 to establish the connection and provide transport for the tunnel. 69.68.125.6 (talk) 13:35, 4 August 2010 (UTC)

SSL/TLS marked as layer 4 (Transport) protocols
(First wiki anything... So go easy!) The OSI Model table at the top of the page indicates that SSL/TLS are transport layer protocols. This is not the case, can they be moved up to the Application layer? OzBluBoy (talk) 14:57, 18 April 2010 (UTC)

You're almost right - they belong to the Presentation layer (i.e. cyrpto) - This is where they were in the Examples table so I've changed it to maintain consistency (and accuracy)

You're both wrong, it belongs in layer 5, that is why it is called Transport Layer Security, it secures the Transport layer below it. Although, technically, it's part of the TCP/IP model which sometimes doesn't map too well when something is "Layer 4" in that stack. — Preceding unsigned comment added by 128.229.4.2 (talk) 14:04, 20 July 2011 (UTC)

Misconceptions
I am currently studying the OSI standard documentation from the ISO (as linked at the end of the article) and have found a number of things that are different from what I have read and heard elsewhere. I am going to start a list here and when I have time I will make corrections in the article.


 * Layer 7 consists of everything above layer 6 (7.1.2.1).
 * Application programs do not interact with layer 7, they are part of layer 7 (7.1.4.1).
 * The user ("human" in the documentation) is also part of layer 7 (7.1.4.1).
 * All of the functions of the Presentation Layer involve syntax (7.2.3.1). Encryption is not mentioned anywhere in the documentation.
 * The Session Layer specifically does not establish checkpointing associated with synchronization (7.3.3.7.3).
 * The physical media is not part of layer 1 but is below layer 1 (6.1.3).

This list is not complete and I will add to it later as needed. Rsduhamel (talk) 14:50, 23 April 2010 (UTC)

Actually, it may be a while before I get a chance to put any of this into the article. If someone else wants to double-check me and put it in, be my guest. Rsduhamel (talk) 17:58, 3 May 2010 (UTC)

Main links
Recent edit comment by, "Main links not needed when the link is at beginning of first sentence". Is this in the WP:MOS somewhere? --Kvng (talk) 22:51, 26 August 2010 (UTC)
 * I haven't seen such a MOS statement, but it seems reasonable in this case. I think the article does not benefit from seven redundant "main" links. Johnuniq (talk) 03:08, 27 August 2010 (UTC)
 * I'm not contemplating reverting the edits, just trying to learn to be a better editor. --Kvng (talk) 14:21, 27 August 2010 (UTC)
 * WP:SUMMARY is the closest MOS entry I've found related to this. Do we consider this to be a summary article? --Kvng (talk) 17:39, 27 August 2010 (UTC)

SIP -- both Layer 7 and Layer 5?
The Article includes SIP in both the Application Layer (Layer 7) and the Session Layer (Layer 5), both in the table and in the right-hand-side box. Is this and error or was this deliberate? The Wikipedia page for SIP and RFC 3261 both state that SIP is an Application-layer protocol. Rahul (talk) 07:47, 22 November 2010 (UTC)

SPDY in session layer??
SPDY is listed in the session layer. But the main article for SPDY describes it as an application-level protocol (and it solves the same problems as HTTP/HTTPS, which are app-level). Does someone know if its placement in session layer is correct? — Preceding unsigned comment added by 98.217.176.221 (talk) 17:28, 28 May 2011 (UTC)


 * Corrected. -- Tomtefarbror (talk) 16:51, 28 August 2011 (UTC)

Internet Engineering Task Force (IETF)
Many questions raised in the Discussion would be answered by including a column in the Examples table, for IETF RFCs. An IETF columns belongs in it, anyway.

FWIW IETF is the source of information used for writing ANSI specs. ANSI determines what the U.S. contributes to ISO specs. Kernel.package (talk) 18:58, 14 June 2011 (UTC)

Edit request from Scottghall, 20 July 2011
The graphics in the can be more legible by using this graphic:

http://bcamca.in/wp-content/uploads/2011/04/osi-model.png

Scottghall (talk) 22:08, 20 July 2011 (UTC)
 * Pictogram voting question.svg Question: Does that image have a copyright? If so it can be uploaded to the commons and then requested again to be put into the article.  See Uploading images for more information. Jnorton7558 (talk) 23:39, 20 July 2011 (UTC)

Two devices in protocol sidebar
In the sidebar, "hub" is listed as Layer 1. While a hub is a device that operates at Layer 1, I don't believe physical devices should be listed on the table. Likewise for Network Switch listed on Layer 2.

If devices are going to be listed, where does it end, pots is listed, should we then list all the devices that pots plugs into(DMS, 5E, Telica, PBX, etc...) I don't think that's the route the article should go, let the Ethernet Hub, Ethernet Switch pages describe there functionality in reguards to the Protocol they run, IE 802.2/802.3/Ethernet II, etc....

— Preceding unsigned comment added by KMurphy111 (talk • contribs) 15:26, July 12, 2012


 * Good point! The right place to discuss this would be Template_talk:OSIstack, the discussion page of the sidebar. You may want to post you request there. I'll support it. Cheers, --EnOreg (talk) 23:19, 12 July 2012 (UTC)


 * I disagree. The sidebar may not be complete, but that's not it's purpose here. It only lists a few popular examples for each layer, so one can get an understanding of the model. One could perhaps annotate that it is not complete, since it seems that this is not obvious. --GGShinobi (talk) 11:15, 15 July 2012 (UTC)


 * Again, please take this discussion where it belongs.--EnOreg (talk) 17:34, 15 July 2012 (UTC)

Unhelpful help
Section Layer 1: Physical Layer, second paragraph – the paragraph actually confuses more than it clarifies. The trouble is that the reasoning "think of Layer 1 as 1 connection, but Layer 2 as 2 or more connections" confuses numerics with meta thinking – it's more informative to explain that Layer 2 peruses and remedies problems occuring in Layer 1. I don't like the paragraph. The section is better without it. Rursus dixit. ( m bork3 !) 08:33, 27 July 2011 (UTC)

Capitalized layers
Hi, I don't see why the layers (layer 1, physical layer etc.) should be capitalized. While they certainly are established concepts they are not proper names. I would like to correct that. Thanks, --EnOreg (talk) 10:14, 13 September 2011 (UTC)
 * If you don't see why, then please study the subject matter and the defining documents, and don't presume to just come here and changes things out of ignorance. Too many people come to WP and want to make changes to things they don't understand quite obviously. Kbrose (talk)


 * Yes, they are proper nouns: they are the names of the OSI layers. They are used in this article to refer to the specific layers of the OSI protocol and no others. — Dgtsyb (talk) 18:09, 13 September 2011 (UTC)
 * But they are not themselves the protocol. The protocol probably qualifies to be a proper noun, and therefore to be upcased, but the generic components of such a protocol (used in other protocols, I presume) are like "computer": downcased. There is every reason not to upcase "physical layer" and "data link layer". Otherwise, the whole point of upcasing gets watered down and starts to be meaningless to our readers. BTW, I am surprised that the "m" in the article title isn't upcased, since it's a proprietary name, isn't it. And the opening line is weird: "The Open Systems Interconnection model (OSI model) is a product of the Open Systems Interconnection effort ...". Surely "Model" is part of the name. If it were "An Open Systems Interconnection model", sure, you'd use "m", or possibly downcase the entire item (i.e., one of many such models, or so widespread that it has lost the ownership/specificity that underpinned the original capitals, like hoovering the carpet, not Hoovering the carpet). BTW, why is "X.200" boldfaced? PS User:EnOreg pointed me here, since we've been discussing capitalisation in this area, and I've been looking at the sorry state of the MoS subpage on this field. Tony   (talk)  02:22, 14 September 2011 (UTC)
 * You presume wrong: they are not used in other models. Referring to "layer 7" has no meaning in this context.  OTOH, "Layer 7" is the "Application Layer".  You will not see any lowercasing even in the talk page, (except here).  Also, they are capitalized in X.200 even though ITU is famous for lower-casing nouns that they coined themselves. — Dgtsyb (talk) 03:56, 14 September 2011 (UTC)
 * As far as OSI model is concerned, ITU refers to it in X.200 as the "Reference Model" or the "Basic Reference Model", capitalized. So that should probably be changed to "OSI Reference Model". — Dgtsyb (talk) 04:00, 14 September 2011 (UTC)
 * The layer names are most certainly proper names, they are the defined names of these layers in OSI, alongside with the numbered versions. And no, model is not part of the name of the concept. It is a model with the adjective OSI but not more. It is very tiresome to see inexperienced editors in these subject matters come every few months or years and try to change or question what has been hashed over many times. Simply, if you don't understand the subject matter, just stay away, you are causing more waste of time than anything else. Kbrose (talk) 04:05, 14 September 2011 (UTC)


 * Kbrose, sorry, you're showing distinct signs of WP:OWNERSHIP, but this is not going to go away, so I ask that you treat other editors with a little more dignity. You can feel it's tiresome for as long as you like, but please hold the insults about inexperienced editors. On your specific point, you haven't convinced me that "model" isn't part of the proper noun: there's only one of these, isn't there? It's unique, and a title. It's odd that you think on the other hand that the numbered layers need a cap (I don't disagree). Dgtsyb, it's generally considered polite not to start with "You presumed wrong": you also seem to be showing early signs of ownership. When you say, "You will not see any lowercasing even in the talk page", that is more likely to make me draw the opposite conclusion. Please remember that WP's readership is much wider than the experts and professionals in a particular area, and that maximising the use of caps for no good reason aside from the fact that people on this talk page might do it, and some publications haven't thought it through, needs to be questioned. On the specifics, "Layer 7", yes, caps there would be endorsed by a wide range of style guides, and WP's own house style (although many would be happy with the lowercase option). But "physical layer" is not unique to this model, surely? The distinction needs to be made, if so. Tony   (talk)  04:13, 14 September 2011 (UTC)
 * The "Physical Layer" is unique to this model, which all experts in the subject matter understand. — Dgtsyb (talk) 19:14, 14 September 2011 (UTC)
 * These things have nothing to do with ownership, but knowledge and experience. Again, you are showing you are lacking such and should be content with reading and learning not editing. Many systems have physical layers, but when the subject is the Physical Layer, it is the instance of such a concept in OSI. TCP/IP on the other hand doesn't have a Physical Layer definition, but it is assumes a physical layer underneath the model. These usages are really completely covered by standard English rules for common and proper nouns.  A common problem on WP occurs when people take details from technical reference works in which certain terms are defined as proper names with definitive meaning in the framework of the specification and write self-standing articles without properly specifying or building the context within which these terms are in fact proper nouns. Kbrose (talk) 04:26, 14 September 2011 (UTC)
 * I am not accusing you of ignorance in English-language style, or sloppiness (even though you don't seem to notice the inadequacies of the article in those respects). So please don't accuse me: "Again, you are showing you are lacking such and should be content with reading and learning not editing" and the extraordinary, "just stay away, you are causing more waste of time than anything else". They are rude, gratuitous insults that border on the xenophobic, and the more you persist, the more you'll see of me. You need to take a check of your attitude, and you'll also need to get used to justifying upper-case usage where there's doubt. I do appreciate your expertise, BTW: thank you for explaining. But I'm not yet sure that I accept your conclusions: if many systems have physical layers, why is this particular physical layer a Physical Layer? Why do you say "model" should be in lower case in the title, then? It just doesn't seem logical. And I note that the first caption has "OSI-Model" and "layers 3 to 5": huh? And "Example" in the middle, with a wrong upper-case E. You can imagine why other editors are starting to ask questions. There are other stylistic glitches littered through the article, too; for example, why do we have "Service Data Unit (SDU)" but then "protocol data unit (PDU)", in the same sentence, if you please? If the articles in this field are so sloppy and inconsistent, why are you spinning out when people come in to assist? Sounds like it's a certain advantage not to be in the field, in that respect, but expert–outside collaboration is clearly needed. So, again, cut the rudeness and engage with us civilly, please. Tony   (talk)  04:41, 14 September 2011 (UTC)
 * These issues in capitalization results from people that know nothing of the context of the usage coming in an changing capitalization here and there from time to time, like now, destroying the consistent usage in the entire article. If you would like to clean it up, Capitalizing Everything would be a better start that lower-casing everything. — Dgtsyb (talk) 19:17, 14 September 2011 (UTC)

I'll try to summarize the arguments we have so far on whether the layers are proper nouns. I would argue as follows. The actual OSI protocols were a failure—they never played any practical role in the industry. But the reference model caught on and has been widely used to classify any protocol into one of the layers. Therefore, the layers have become generic networking terms that do not qualify as proper nouns.
 * Proper nouns:
 * (1) They are the names of the OSI layers, referring to the specific layers of the OSI protocol and no others. E.g., the "Physical Layer" is unique to this model.
 * (2) They are capitalized in X.200 even though ITU is famous for lower-casing nouns that they coined themselves.
 * Not proper nouns:
 * (3) The layers are generic and applied to other protocols.

A Google search produces corroborating evidence, suggesting that the vast majority of university lectures and research papers spell the layers in lower case. The same is true for the educational material by one of the largest vendors, Cisco. Even more strikingly, the standard textbooks on the subject, most prominently Tanenbaum as well as Peterson & Davie, use the lower case. Unlike standards documents, textbooks undergo rigorous copy-editing by language experts. All this seems to make a strong case for lower-case layers. --EnOreg (talk) 14:05, 17 September 2011 (UTC)


 * Wow! Are you saying that the relative failure or adoption rate of an ISO standard determines whether the nouns in WP articles are proper or not?  I didn't see that reasoning in WP:CAPS. ;) — Dgtsyb (talk) 01:30, 19 September 2011 (UTC)


 * Nah, that is mostly background information. Please consider the entire argument. --EnOreg (talk) 09:21, 21 September 2011 (UTC)

Are there any other arguments we're missing? Or are the arguments I laid out above convincing? --EnOreg (talk) 15:37, 28 September 2011 (UTC)
 * If there are no more objections let's change the layers to lower case. --EnOreg (talk) 10:19, 3 October 2011 (UTC)
 * I object: they are proper nouns. — Dgtsyb (talk) 18:12, 3 October 2011 (UTC)
 * The layer names are clearly proper nouns, they are the given names of the defining characteristics of the standard. You name your son John or Shorty, you capitalize it. The Moon is our companion, despite many other planets having a moon, too. Kbrose (talk) 00:11, 4 October 2011 (UTC)


 * Then let's hear your arguments. I've presented mine on September 17, countering the previous ones for proper nouns. How do you argue against these? --EnOreg (talk) 08:38, 5 October 2011 (UTC)

It's been a month since the arguments for proper nouns have been refuted and no new ones have been presented since. That leaves really no reason not to adapt the article to the spelling commonly found in the literature on the subject. --EnOreg (talk) 15:42, 16 October 2011 (UTC)


 * The arguments were not refuted. They are proper nouns and must stand. — Dgtsyb (talk) 02:30, 17 October 2011 (UTC)


 * On September 17 I summarized all arguments pro caps I could find in this discussion and presented arguments that I think refute both of them. Since then no counter-arguments have been presented. If you think the pro arguments were not refuted you need to say either which pro argument I missed, or which part of my refutation failed to convince you and why. Only repeating your conclusion doesn't count. You need to reveal your arguments to have a grown-up discussion. --EnOreg (talk) 20:21, 17 October 2011 (UTC)

Having not read all the above, I simply consulted books, and found that the layer "names" are not typically treated as proper names:, , , generic; ,  capitalize "Application" etc. but not "layer"; in the first page of 10 google book hits for OSI and "Application Layer", exactly zero show evidence of treatment of "Application Layer" as a proper name.  For OSI and "Physical Layer", did find one out of ten:  (it also includes "The Lemon-Pudding Layer", so clearly a great moist source). Dicklyon (talk) 04:22, 18 October 2011 (UTC)

Lacking any credible support for the assertion "they are proper names", and lots of evidence of generic lowercase usage in reliable sources about the OSI model, I think it's time for a multiple RM on the various capitalized "Layer" articles, and corresponding fixes to this one. EnOreg, you want to start that, or should I when I have time to read WP:RM about to do a multiple move request? Dicklyon (talk) 19:49, 18 October 2011 (UTC)

OK, I have gone ahead and started a multiple RM at Talk:Physical Layer. Dicklyon (talk) 03:46, 19 October 2011 (UTC)


 * Thanks Dicklyon. --EnOreg (talk) 07:34, 19 October 2011 (UTC)
 * Odd that nobody has bothered to oppose the multiple move, given all the assertions above. Dicklyon (talk) 23:20, 25 October 2011 (UTC)
 * Since the moved happened, I went ahead and did a first-pass case correction here. If I got anything wrong, please tweak it.  Dicklyon (talk) 18:57, 26 October 2011 (UTC)

PPTP is session layer or data link layer??? Please fix contradiction
If you look at the table at top right corner of the article, PPTP belongs to "session layer". If you look down in the table with examples, PPTP belongs to the "data link layer". I am not sure, where it belongs more, but please fix it ASAP. --Sena (talk) 08:49, 14 September 2011 (UTC)
 * Well, it seems that PPTP belongs to the "data link layer". You can find it in this table OSI_model and in lots of places in internet. If nobody will object it, I will move PPTP to the data link level in 7 days. --Sena (talk) 08:31, 15 September 2011 (UTC)
 * No, it is not listed as at the link layer in OSI, only for TCP/IP as in the comparison in the table. In pure OSI it literally cannot be below a connection-oriented layer, e.g. Transport Layer, since it requires the services of TCP. Please make no changes out of ignorance or what other dubious sources may lead you to conclude. Kbrose (talk) 15:44, 15 September 2011 (UTC)
 * Ah, OK. Thanks for explanations. Now I get it. PPTP is on the "data link" level for the TCP/IP "mapped" on OSI model. So, in the top-right table it is shown in pure OSI model. Great, but I think this must be somehow better explained. First quick look shows some kind of contradiction. --Sena (talk) 16:45, 15 September 2011 (UTC)

7 layer model
I hear many people calling this model the 7 layer model, could someone please redirect that term (seven layer model) to this page? — Preceding unsigned comment added by 89.138.50.78 (talk) 10:03, 10 October 2011 (UTC)
 * These exist: Seven-layer model, Seven-layer OSI model. Dicklyon (talk) 19:43, 18 October 2011 (UTC)

datagram service model?
In the section OSI model, the description of Multiprotocol Label Switching (MPLS) states, "It was designed to provide a unified data-carrying service for both circuit-based clients and packet-switching clients which provide a datagram service model." There are two ambiguities with this: Will an MPLS expert please revise this? Stephen Charles Thompson (talk) 21:19, 24 January 2012 (UTC)
 * 1) Dangling participle error: What provides the datagram service model?  Is the provider one or more of the clients, the data carrying service, or the MPLS function being described?
 * 2) What is a datagram service model?  That phrase does not appear in any other articles and it is not defined in context.

Layer 0
There is now Layer 0 which is seen on ROADM (Reconfigurable Optical Add/Drop Multiplexor) Fujitsu flashwaves do layer 0, as does Cienna CPL. — Preceding unsigned comment added by 50.98.213.201 (talk) 21:25, 21 April 2012 (UTC)

Incorrect Link on 3rd graphic
On 3rd graphic (Data Unit -- Layer -- Function), on 2nd Layer "Data Link", on Function column, there is a link to "Physical Layer" that points to a Wikipedia article about "Memory Physical directions", and not "Network Physical directions". I thinks it's more correct to point it to: http://en.wikipedia.org/wiki/MAC_address  — Preceding unsigned comment added by 186.130.41.69 (talk) 15:04, 20 June 2012 (UTC)
 * Thanks, I fixed that. Johnuniq (talk) 02:16, 21 June 2012 (UTC)

errors in description of Layer 1
According to Tanenbaum ("Computer Networks") and as far as I remember, flow control is not part of the physical layer (1), but of the data link layer (2), and also occurs to some degree within the network layer (3). Also, the physical layer has nothing to do with the sharing of resources among multiple users. That's a task of layer 2 (MAC). Layer 1 covers mainly physical concerns (thus the name) like the media (cable, radio, ...) and how a bit is represented there. --GGShinobi (talk) 21:19, 9 July 2012 (UTC)

Layer 3: network layer - Address not hierarchical?
In the layer 3 section, there is the phrase "The addressing scheme is not hierarchical.". From my basic understanding of networking, layer 3 addressing IS hierarchical, in contrast to layer 2 one.

Regards Francesco — Preceding unsigned comment added by Francesco87 (talk • contribs) 20:16, 22 October 2012 (UTC)

"History" - what about a date?
The "history" section doesn't mention even a single date... — Preceding unsigned comment added by 80.174.219.223 (talk) 16:58, 14 November 2011 (UTC)

I came here to say that, what use is a history (an ordering of events) that doesn't provide any dates to put it into context with other contemporaneous events. — Preceding unsigned comment added by 27.252.100.236 (talk) 12:37, 7 August 2013 (UTC)
 * Egad, the current article is quite wacky. Generally unsourced, and seems mostly to be a vandalism magnet. Article mentions Charles Bachman, but he was a database person, not networking. The reality is the late 1970s and early 1980s work by Hubert Zimmermann and John Day (computer scientist) which needs to be clarified with citations. Update: it appears that Bachman did play a role for a time as chair of the ISO/TC97/SC16 between 1977 and 1982. See this source for example, W Nowicki (talk) 18:38, 3 September 2013 (UTC)

"History" - OSI vs. TCP/IP
Also, was OSI ever implemented, and what is the current state of that implementation (if any)? The intro para calls it a "conceptual model," and there's no mention in the History sxn of an actual implementation. My mid-understanding is that TCP/IP became the de facto standard instead. There's some discussion of this on pg 457 of "The dream machine: J.C.R. Licklider and the revolution that made computing personal" by M. Mitchell Waldrop. The discussion makes it sound like OSI (or perhaps support for OSI) was in part a European political response to the fact that TCP/IP was American (not European) and had been developed under ARPA (therefore US DOD) funding. That of course means this question of why TCP/IP is still around, and OSI isn't (if that's correct), is a contentious issue. I don't know how to evaluate that claim. Mcswell (talk) 14:02, 29 May 2013 (UTC)
 * Alas, much of this was personal anecdote and thus not really acceptable for Wikipedia. However, this article from July 2013 would be a good one to cite. And of course needs to clarify the model vs. a specific protocol stack. W Nowicki (talk) 18:07, 3 September 2013 (UTC)

The information in parentheses (ISO/IEC 7498-1) tells you unambiguously that "The Open Systems Interconnection (OSI) model (ISO/IEC 7498-1)" is an ISO/IEC International Standard. For precision in labeling, that label should include the date (ISO/IEC 7498-1:1994), but the absence of the date does not change its status. TCP/IP is one implementation of two of the layers of that standard. Please read for yourself about the labeling that the ISO uses to specify the status of a proposed standard as it works its way through the standardization process at: https://en.wikipedia.org/wiki/International_Organization_for_Standardization Wikifan2744 (talk) 02:25, 9 September 2013 (UTC)

I withdraw my sentence just above: "TCP/IP is one implementation of two of the layers of that standard." It's much more complicated than that, far more complicated than I can explain. The Internet protocol suite, which includes the TCP/IP networking model and related protocols (TCP, IP, and others), was developed by and is maintained by the Internet Engineering Task Force (IETF): https://en.wikipedia.org/wiki/Internet_Engineering_Task_Force. Note that there are both the model and also the protocols. Models and protocols are separate concepts that are easily conflated or confused in writing and speaking.

The article on the IETF begins as follows: "The Internet Engineering Task Force (IETF) develops and promotes Internet standards, cooperating closely with the W3C and ISO/IEC standards bodies and dealing in particular with standards of the Internet protocol suite (TCP/IP).[1][2] It is an open standards organization, with no formal membership or membership requirements."

The article on the Internet protocol suite (commonly known as TCP/IP) is long and comprehensive. https://en.wikipedia.org/wiki/Internet_protocol_suite  Its opening paragraph begins: "The Internet protocol suite is the networking model and a set of communications protocols used for the Internet and similar networks. It is commonly known as TCP/IP, because its most important protocols, the Transmission Control Protocol (TCP) and the Internet Protocol (IP) were the first networking protocols defined in this standard...." Please read the article for further details, but pertinent to this discussion is the section just below.

"OSI and TCP/IP layering differences[edit source | editbeta]

"The three top layers in the OSI model—the application layer, the presentation layer and the session layer—are not distinguished separately in the TCP/IP model where it is just the application layer. While some pure OSI protocol applications, such as X.400, also combined them, there is no requirement that a TCP/IP protocol stack must impose monolithic architecture above the transport layer. For example, the NFS application protocol runs over the eXternal Data Representation (XDR) presentation protocol, which, in turn, runs over a protocol called Remote Procedure Call (RPC). RPC provides reliable record transmission, so it can safely use the best-effort UDP transport.

"Different authors have interpreted the RFCs differently, about whether the link layer (and the TCP/IP model) covers OSI model layer 1 (physical layer) issues, or whether a hardware layer is assumed below the link layer.

"Several authors have attempted to incorporate the OSI model's layers 1 and 2 into the TCP/IP model, since these are commonly referred to in modern standards (for example, by IEEE and ITU). This often results in a model with five layers, where the link layer or network access layer is split into the OSI model's layers 1 and 2.

"The session layer roughly corresponds to the Telnet virtual terminal functionality [citation needed], which is part of text based protocols such as the HTTP and SMTP TCP/IP model application layer protocols. It also corresponds to TCP and UDP port numbering, which is considered as part of the transport layer in the TCP/IP model. Some functions that would have been performed by an OSI presentation layer are realized at the Internet application layer using the MIME standard, which is used in application layer protocols such as HTTP and SMTP.

"The IETF protocol development effort is not concerned with strict layering. Some of its protocols may not fit cleanly into the OSI model, although RFCs sometimes refer to it and often use the old OSI layer numbers. The IETF has repeatedly stated [citation needed] that Internet protocol and architecture development is not intended to be OSI-compliant. RFC 3439, addressing Internet architecture, contains a section entitled: "Layering Considered Harmful".[28]

"Conflicts are apparent also in the original OSI model, ISO 7498, when not considering the annexes to this model (e.g., ISO 7498/4 Management Framework), or the ISO 8648 Internal Organization of the Network layer (IONL). When the IONL and Management Framework documents are considered, the ICMP and IGMP are neatly defined as layer management protocols for the network layer. In like manner, the IONL provides a structure for "subnetwork dependent convergence facilities" such as ARP and RARP.

"IETF protocols can be encapsulated recursively, as demonstrated by tunneling protocols such as Generic Routing Encapsulation (GRE). GRE uses the same mechanism that OSI uses for tunneling at the network layer."

Wikifan2744 (talk) 06:29, 9 September 2013 (UTC)

Semi-protected edit request on 31 July 2014
DHCP is a network layer protocol and not application layer. these are all for delivering IP addresses to the node including other information.

194.213.3.4 (talk) 13:18, 31 July 2014 (UTC)


 * No, DHCP messages are ultimately sent with IP, so that can't be right.  Anon 126   (notify me of responses! / talk / contribs) 21:07, 9 August 2014 (UTC)

Inception
This indicates that the inception of the OSI model was 1984. I'm not sure what an inception is. ~KvnG 18:42, 6 October 2014 (UTC)

IP/OSI confusion in layer box
The box of networking layers here makes the mistake of conflating TCP/IP network protocols with the OSI layers. This is wrong; the two are not connected, not even remotely.

I realize that many educational institutions like to use the OSI model because it's such a neat network model and it's so nice to explain things, but it's 'wrong' to assume that the IP network protocol is situated in layer 4 of the OSI model, or that the SIP protocol is situated in layer 7 of the OSI model. They are not. The IP protocol is found in the network layer of the TCP/IP "model", and the SIP protocol is found in the applicatoin layer of the TCP/IP "model".

If you're going to show the OSI model, the protocols on OSI protocols should be used instead. — Preceding unsigned comment added by 84.197.94.196 (talk) 23:03, 19 April 2013 (UTC)


 * The OSI model does not only apply to the OSI protocols as the networking literature attests. --EnOreg (talk) 15:26, 21 April 2013 (UTC)


 * There are no "OSI protocols." The OSI model doesn't define any protocols. By definition, protocols such as TCP/IP fall within the scope of the OSI model. The official documentation says:


 * "The purpose of this Reference Model of Open Systems Interconnection is to provide a common basis for the coordination of standards development for the purpose of systems interconnection, while allowing existing standards to be placed into perspective within the overall Reference Model."


 * The tasks generally performed in the Internet layer of the TCP/IP model fall within the general description of the Network layer of the OSI model. Rsduhamel (talk) 23:44, 26 January 2015 (UTC)

Examples wrong
the table in Examples gives an entry for "Layer 2" X "TCP/IP protocols". This is wrong. Layer 2 protocols, like PPP, is at a level where IP doesn't exist (IP is on top of L2).

Zenkutsu (talk) 18:42, 9 February 2015 (UTC)

Also things like HTTP and HTML are backwards. HTML is transferred over HTTP, so HTML should be layer 7, not layer 6, and HTTP should be layer 6 not layer 7. The same goes for the other 6/7 layer examples.

Majenko (talk) 23:58, 14 December 2015 (UTC)

Image formats and styling languages in layer 6?
In the table in section "Description of OSI layers" there are some interesting examples for layer 6: HTML, CSS, GIF. How so? Can somebody explain? What does an image format have to do with the presentation layer of the OSI model? And HTML/CSS? This looks like it was mistaken for "visual presentation of a web page". Similarly SQL is a weird choice for layer 5 because that's a programming language, not a protocol. — Preceding unsigned comment added by Niko IstDerNameImmerNochZuÄhnlich (talk • contribs) 01:35, 5 October 2015 (UTC)

It seems to me that the examples in the upper layers are all completely mixed up. HTML should be a higher layer than HTTP since HTML is transferred using the HTTP protocol over a TCP socket using IP. The 6/7 layer examples are all backwards.

Majenko (talk) 00:00, 15 December 2015 (UTC)

Layer 6: Presentation Layer - "big mapping"?
In first paragraph of Layer 6 topic, should "big mapping" be changed to "bit mapping"? Thank you. — Preceding unsigned comment added by 198.28.128.150 (talk) 16:53, 20 May 2016 (UTC)


 * No, just to "mapping". I've done so. Guy Harris (talk) 17:24, 20 May 2016 (UTC)

Layer 5 (Session) examples incorrect?
Numerous text-books and online sources use HTTP; FTP; RDP; SSH; Telnet as examples for Layer 7 (Application).

Even other wiki pages like List of network protocols (OSI model) and Application Layer — Preceding unsigned comment added by Beanaroo (talk • contribs) 09:48, 23 March 2016 (UTC)


 * Every one of the examples you have listed is an IP model protocol. To quote ISO/IEC 7498-1, which defines the OSI model, "This reference model provides a common basis for the coordination of standards development"; the development of HTTP, FTP, RDP, SSH, and Telnet was NOT coordinated b the OSI model, and did NOT use the OSI model as a basis; instead, the 5-layer IP model model was used when developing these protocols.  In IP model terms, layer 4 is the "internetwork" layer, and layer 5 is the "application" layer.  Application protocols designed using the IP model as a basis that need session management have to do it built-in, because using the IP model as a basis leads to not having a separate session management protocol.  On the other hand, application protocols designed using the OSI model as a basis can outsource it to a separate protocol.  Because of this, when using the OSI model to describe IP model protocols, we can say that a protocol spans multiple layers; different parts of the protocol fill the roles in different layers. ~ LukeShu (talk) 07:43, 8 December 2016 (UTC)

PHY definition
The section describing the PHY layer states: "It defines the network topology as bus, mesh, or ring being some of the most common." This is not correct. E.g. there exists both WiFi star and mesh network implementations based on the same PHY's. Another example with radio PHY's you may implement star, mesh, ad-hoc networks Cebmeisner (talk) 12:56, 15 May 2017 (UTC)


 * So update it to discuss radio PHYs. (The two examples you give are both radio.) Guy Harris (talk) 18:55, 15 May 2017 (UTC)

Article is inconsistent on whether the presentation layer include compression and encryption
The Layer 6: Presentation Layer subtopic claims (without citation) that it is a myth that the presentation layer does compression and encryption. However, the table at the top of Description of OSI Layers mentions compression and encryption as layer 6 functions. Which is it? — Preceding unsigned comment added by Marcelocantos (talk • contribs) 23:05, 13 October 2017 (UTC)


 * The claim that it's a myth that compression can be done by the presentation layer is completely wrong for compression - ISO/IEC 7498-1:1994 (at least in its available-for-free X.200 version) explicitly says:


 * 7.2.4 Functions within the Presentation Layer
 * The Presentation Layer performs the following functions to help accomplish the presentation-services:
 * a) negotiation and re-negotiation of transfer syntax;
 * b) representation of the abstract syntax chosen by the application-entities in the transfer syntax negotiated or renegotiated, including format and special purpose transformations (for example, data compression)
 * a) negotiation and re-negotiation of transfer syntax;
 * b) representation of the abstract syntax chosen by the application-entities in the transfer syntax negotiated or renegotiated, including format and special purpose transformations (for example, data compression)
 * b) representation of the abstract syntax chosen by the application-entities in the transfer syntax negotiated or renegotiated, including format and special purpose transformations (for example, data compression)


 * (emphasis mine). I'm not seeing anything explicit about encryption, but encryption sure sounds like a "special purpose transformation" to me.  I've remove the claim about compression and, if I can find a reference, I'll rip out the claim about encryption as well. Guy Harris (talk) 00:14, 14 October 2017 (UTC)


 * Oh, to heck with it. I've just removed the whole "myth" claim, as, given that the "compression" part is explicitly proven wrong by the actual OSI reference model document, I see no reason to believe the "encryption" part. Guy Harris (talk) 00:18, 14 October 2017 (UTC)


 * X.200 (1988) also includes compression, although it also says "Recommendation X.200 and ISO 7498 [Information Processing Systems – Open Systems Interconnection – Basic Reference Model] were developed in close collaboration and are technically aligned.", so maybe they were "technically aligned" but the CCITT added compression themselves; I don't have the older version of the ISO spec handy. Perhaps the original spec didn't include it, but it was apparently added subsequently. Guy Harris (talk) 00:41, 14 October 2017 (UTC)
 * The confusion may be due to the fact that the OSI model, while apparently implemented, is irrelevant as far as the internet goes, apart from its use for interrogating people taking various network certification tests. My understanding (very limited) is that the presentation layer does nothing in practical systems, and that might have been behind the original research regarding "common myth". I think the OSI designers intended the presentation layer to handle things like encryption and compression, as well as EBCDIC/ASCII translation as mentioned in the article. Johnuniq (talk) 00:57, 14 October 2017 (UTC)

Defining a standard
The article says in several places that the OSI model defines a standard. The official documentation specifically says it does not define a standard. Rsduhamel (talk) 16:46, 19 January 2015 (UTC)
 * But ISO 7498 says right on its front cover that it's a standard. Please explain. --Wtshymanski (talk) 22:07, 26 January 2015 (UTC)


 * The cover page of the ISO version and page 1 of both versions are titled "International Standard" (from page 1 on, both the ISO and ITU versions are identical). However, the body of the document is clear that the document does not set a standard. Specifically, it says:


 * "It is not the intent of this Reference Model either to serve as an implementation specification, or to be a basis for appraising the conformance of actual implementations, or to provide a sufficient level of detail to define precisely the services and protocols of the interconnection architecture. Rather, this Reference Model provides a conceptual and functional framework which allows international teams of experts to work productively and independently on the development of standards for each layer of the Reference Model for OSI."


 * Although it does not say "this is not a standard" it is described as a reference model and framework to allow the development of standards. Perhaps I am being imprecise but I have seen a lot written on the Internet and in textbooks describing the OSI model as a standard that must be complied with. I've also seen it described as an attempt by bubble-headed European bureaucrats to shove a restrictive, complicated and unworkable standard down the throats of developers. It is no such thing. It is just framework describing what tasks may or may not be done by networking applications. Developers don't necessarily follow the OSI model. The OSI model follows what developers generally do. Rsduhamel (talk) 23:29, 26 January 2015 (UTC)


 * This is really moot, because the OSI model is really used nowhere as a development model, at least not above Layer 2. The lack of usable protocols in the late 80s, early 90s, made it obsolete quickly and TCP/IP has reigned, despite its shortcomings. I don't see a strict contraction in a reference model being a standard. It certainly is a standard way of teaching network application architecture. Developers use the Internet Protocol Suite (TCP/IP) as their model. Anything the standards organizations deal with is codified eventually as a standard, seems that's the only way to document their work. Kbrose (talk) 14:04, 27 January 2015 (UTC)


 * Call it a standard if you will. There is no point in arguing semantics when the article has much more serious issues. Someone who has actually read the specification needs to give it some serious attention, if it's worth the effort. Rsduhamel (talk) 02:50, 3 February 2015 (UTC)


 * The whole problem is, even if there is a specification, in a general article such as this one it is sort of futile to attempt to rewrite it. People can go to the spec for an official read.  OSI is theoretical not practical.  It is not useful to explain it from the detail of an implementation perspective because it is never implemented -- the world has chosen the TCP/IP five layer model, and anyone reading Wikipedia to learn about networking should not waste a lot of time reading about OSI (They will have plenty of time to do this if they ever try to hack an OSI implementation such as HL7.).  Therefore, this article should be dramatically simplified.


 * It is misleading because this page is referenced from pages that discuss very important practical network topics.Squeeky Longhair (talk) 17:57, 9 February 2018 (UTC)

history
according to the interview with Charles Bachman IBMs SNA was one of the blueprints for the OSI model

"I started looking around for useful solutions to similar problems and became familiar with the IBM project called SNA or System Network Architecture. It had some very attractive features. It looked like it was well-architected. It had six architectural layers. It had a principle that each layer only talked to the layer immediately above and to the layer immediately below it. A layer might be modified, or replaced, and its protocol changed, as long as it maintained both its upper interface and its lower interface. And so it was possible to swap out a layer with an improved protocol and a new implementation that would support different things without disrupting how the other layers functioned." http://ethw.org/Oral-History:Charles_Bachman

as well as the experiences with standardisation bachmann made with the IDS project: "Honeywell was still willing to consider the communication add-on to IDS, so I started on that project and called it “Distributed System Interconnection.” And so I took on the project, which I had started in Phoenix, under the name of “Inter-Communicator” https://en.wikipedia.org/wiki/Integrated_Data_Store

http://www.dsbaral.com.np/subject/network-programming/sna — Preceding unsigned comment added by 77.180.0.193 (talk) 22:07, 4 July 2018 (UTC)

IPoE redirects here. WHY?
I was looking for info on I(ternet)P(rotocol)o(ver)E(thernet), and got this page. But this page says absolutely NOTHING about IPoE. So where should a seeker after knowledge go? 58.7.61.232 (talk) 03:35, 10 August 2015 (UTC)


 * Perhaps somebody could restart the IPoE page with this http://www.dslreports.com/r0/download/1468367~39a604b7c30141db72e55620ebf95879/Understanding%20PPPoE%20and%20DHCP_200187.pdf info from Mark Bernstein of Juniper Networks. 58.7.61.232 (talk) 03:57, 10 August 2015 (UTC)


 * IPoE is now a page that discusses the use of IP-over-Ethernet in access networks (there's nothing magical about IP-over-Ethernet in general, making it deserving of its own page; it's only omitting PPPoE in access networks that's interesting), so it no longer redirects here (which is as it should be - there's no reason why it should redirect here). Guy Harris (talk) 00:20, 5 July 2018 (UTC)