Talk:OSI model

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

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

• 
  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#Confusing layers (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)

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

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. —The preceding unsigned comment was added by 64.231.152.31 (talk • 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. —The preceding unsigned comment was added by 216.184.31.184 (talk • 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. —The preceding unsigned comment was added by 213.84.162.175 (talk • 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 — 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 — seemingly like all communication stacks/models — 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.

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#Layers in the Internet protocol suite stack 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)

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

OSI Model

OSI Model

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#OSI_Graphic_-_Osi_model_com.jpg --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 [[1]] 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 [[2]] on this matter on this page. This was removed. Anyone know why? It was really useful for non-engineers to follow.

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)

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:

Sublayer	Function	Representative protocol
Subnetwork independent	Agnostic to underlying layers	CLNP, IP, CONS
Subnetwork dependent convergence	Maps between logical network layer address and medium specific address	ARP
Subnetwork access	Media-dependent (overlaps with L2 in some cases)	LLC, LLC/SNAP

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)

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.