Talk:Open Systems Interconnection Reference Model: Difference between revisions
imported>Greg Woodhouse (categories okay) |
John Leach (talk | contribs) m (Text replacement - "OSI model" to "Open Systems Interconnection Reference Model") |
||
(11 intermediate revisions by 4 users not shown) | |||
Line 1: | Line 1: | ||
{{ | {{subpages}} | ||
| | ==Section on why standards are needed== | ||
| | |||
| | I reworked the language somewhat. While it can debated whether or not the previous language represents a neutral point of view, it certain comes across as polemical in tone. The same point can be made without focusing on motives. [[User:Greg Woodhouse|Greg Woodhouse]] 13:48, 9 April 2007 (CDT) | ||
| | |||
| | :If you have more ideas for this thing please go ahead and add them... I'm trying to come up with more content. Maybe something on the old hardware and protocols, etc? Hmm. --[[User:Eric M Gearhart|Eric M Gearhart]] 13:58, 9 April 2007 (CDT) | ||
| | |||
| | Well, it is certainly true that internetworking requires running the same layer 3/4 protocols over very different transports. The Imtermet would hardly have become as ubiquitous as it is if IP applications didn't run over Ethernet, DSL, cable and dialup lines. I remember a number of years ago that a system administrator at what was then my ISP had frame relay set up to his house! How many people are going to do that so they can have Internet access? [[User:Greg Woodhouse|Greg Woodhouse]] 14:13, 9 April 2007 (CDT) | ||
| | |||
== Meaning of switch == | |||
I changed the link back from electronic switch, because in the context of computer networks, "switch" has a different meaning. An electronic switch could be an electronic component like a transistor or a latch, but a network switch is a computing device, albeit of a special type. Both layer 2 devices (such as bridges) and layer 3 devices (such as routers) have been called switches. Radia Perlman (the inventor of the spanning tree bridge) somewhat sarcastically said that switch is a word meaning "fast" (see her book, ''Interconnections'' for a detailed, and readable, treatment). In fact, the term switch really has less to do with how the device ''works'', but with how it is ''used''. ATM switches are technically layer 3 devices, but are used in much the same way as layer 2 switches. "Fast Ethernet" (100baseT and 1000baseT) is switched, even though Ethernet/802.3 is a layer 2 protocol. (Remember that 10baseT does not require switching.) [[User:Greg Woodhouse|Greg Woodhouse]] 11:36, 12 May 2007 (CDT) | |||
::Greg, I have modified the "switch" link on this article to say "network switch" (which I think is the same as a router anyway). The plain old "switch" link is redirected to the [[Electronic switch]]. There are many links to "switch" from various computer articles that really do need to go to [[Electronic switch]]. So we need a disambiguation page, but I don't know how to make one.[[User:Pat Palmer|Pat Palmer]] 11:53, 12 May 2007 (CDT) | |||
:Not necessarily. If you have to physical networks connected by a switch and then use traceroute (tracert on Windows) to path from a host on one segment to one on the other, you won't see the switch. However, if they're connected by a router, you ''will'' see the router. On the other hands, there are contexts where switches and routers are pretty much the same thing. Part of the problem, and something that really needs to be addressed in this article, is that the Open Systems Interconnection Reference Model is not always a perfect fit for network technology, so thee can be a certain degree of arbitrariness in the assignment of network devices to OSI layers. [[User:Greg Woodhouse|Greg Woodhouse]] 12:03, 12 May 2007 (CDT) | |||
==combined and renamed== | |||
I saw no reason for two short articles rather than one longer one, so I combined these under the name I am most familiar with.[[User:Pat Palmer|Pat Palmer]] 21:36, 12 May 2007 (CDT) | |||
==can we move the standards justification stuff?== | |||
I think the section justifying the need for network standards belongs on the [[computer network]] page, rather than here.[[User:Pat Palmer|Pat Palmer]] 21:36, 12 May 2007 (CDT) | |||
:I think that seems sensible. Maybe something along the lines of "for further discussion, see [[Computer network]]" would be helpful. But, then again, maybe even that is not needed. [[User:Greg Woodhouse|Greg Woodhouse]] 22:25, 12 May 2007 (CDT) | |||
::To get to basics, what is the perceived need here for "justifying" network standards? I could put more on [[Computer networking reference models]], and indeed go before computer networking to such things as the [[Kingsbury Compromise]] of 1913, in which AT&T agreed to give up its telegraphy business in exchange for setting technical standards for nationwide telephony. (At this point, why didn't AT&T change its name to AT?) | |||
::Anyway, the main driver for standards came from telephony and radio, well before computers. Once it was in the international arena, OSI was first driven by telephony, not computer people. | |||
::Radio and telephony people know they must interconnect, so standards are important to them. With computers, however, most vendors stayed proprietary as long as possible, to "lock in" customers to their networks. The abortive attempt, in the 1980s, to force OSI protocols (as distinct from the reference models), failed for many reasons, not least of which is the then US-military-subsidized technical standards for the TCP/IP family were available for the cost of downloading electronic documents, while it cost thousands of dollars to get a set of the hard copy of OSI standards. Vendors often bundled TCP/IP free with UNIX systems, where an OSI stack (e.g., X.400 electronic mail) might cost $200,000. | |||
::In other words, standards per se didn't justify OSI. OSI was a consequence of an environment that was accustomed to standards. That environment, however, could not compete economically with the TCP/IP protocol documentation and reference implementations. [[User:Howard C. Berkowitz|Howard C. Berkowitz]] 04:52, 7 May 2008 (CDT) | |||
== Internet protocol coercion == | |||
Let me start with the disclaimer that I do not believe anyone should be teaching or learning the original OSI 7-layer reference model, defined solely by ISO document 7498, because that definition is obsolete even to ISO. Quite a number of extensions were added, such as the ''OSI Management Framework'', the ''OSI Routeing [sic] Framework'', ''Internal organization of the network layer,'' etc. | |||
I suppose we have to have the article, because so much information and misinformation refers to it. If I must use a layered model, which is not always the right tool, I use the 4-layer Internet model. | |||
Nevertheless, I see something very common here: there is an attempt being made to show where Internet protocols fit into this model, and even what physical devices exist at various layers. There are two basic problems with doing so: the Internet Engineering Task Force consciously and explicitly does its protocol development without any concern for OSI compatibility, and the OSI reference model is an abstraction that defines no physical devices. RFC1812, ''Requirements for IP routers'', is a pragmatic IETF definition, of a type ISO avoided "not to constrain implementers". | |||
I have started [[network reference models]] to explain the relationships of the IETF, OSI, ATM, IEEE and other models, and how there are some decided incompatibilities. How can we avoid, in Citizendium, the perpetuation of OSI myths? | |||
As an aside, I started working on the pre-ISO/ITU work in 1979, with the ANSI DISY project. I was the first engineer hired by the Corporation for Open Systems (COS), an industry group chartered to accelerate the introduction of OSI and ISDN protocols. In retrospect, I can look back at six years of my life, in which I was constantly preaching that "OSI is the answer", and see the fundamental problem: we may have had an answer, but we never had a useful definition of the question. [[User:Howard C. Berkowitz|Howard C. Berkowitz]] 11:14, 6 May 2008 (CDT) |
Latest revision as of 05:21, 8 March 2024
Section on why standards are needed
I reworked the language somewhat. While it can debated whether or not the previous language represents a neutral point of view, it certain comes across as polemical in tone. The same point can be made without focusing on motives. Greg Woodhouse 13:48, 9 April 2007 (CDT)
- If you have more ideas for this thing please go ahead and add them... I'm trying to come up with more content. Maybe something on the old hardware and protocols, etc? Hmm. --Eric M Gearhart 13:58, 9 April 2007 (CDT)
Well, it is certainly true that internetworking requires running the same layer 3/4 protocols over very different transports. The Imtermet would hardly have become as ubiquitous as it is if IP applications didn't run over Ethernet, DSL, cable and dialup lines. I remember a number of years ago that a system administrator at what was then my ISP had frame relay set up to his house! How many people are going to do that so they can have Internet access? Greg Woodhouse 14:13, 9 April 2007 (CDT)
Meaning of switch
I changed the link back from electronic switch, because in the context of computer networks, "switch" has a different meaning. An electronic switch could be an electronic component like a transistor or a latch, but a network switch is a computing device, albeit of a special type. Both layer 2 devices (such as bridges) and layer 3 devices (such as routers) have been called switches. Radia Perlman (the inventor of the spanning tree bridge) somewhat sarcastically said that switch is a word meaning "fast" (see her book, Interconnections for a detailed, and readable, treatment). In fact, the term switch really has less to do with how the device works, but with how it is used. ATM switches are technically layer 3 devices, but are used in much the same way as layer 2 switches. "Fast Ethernet" (100baseT and 1000baseT) is switched, even though Ethernet/802.3 is a layer 2 protocol. (Remember that 10baseT does not require switching.) Greg Woodhouse 11:36, 12 May 2007 (CDT)
- Greg, I have modified the "switch" link on this article to say "network switch" (which I think is the same as a router anyway). The plain old "switch" link is redirected to the Electronic switch. There are many links to "switch" from various computer articles that really do need to go to Electronic switch. So we need a disambiguation page, but I don't know how to make one.Pat Palmer 11:53, 12 May 2007 (CDT)
- Not necessarily. If you have to physical networks connected by a switch and then use traceroute (tracert on Windows) to path from a host on one segment to one on the other, you won't see the switch. However, if they're connected by a router, you will see the router. On the other hands, there are contexts where switches and routers are pretty much the same thing. Part of the problem, and something that really needs to be addressed in this article, is that the Open Systems Interconnection Reference Model is not always a perfect fit for network technology, so thee can be a certain degree of arbitrariness in the assignment of network devices to OSI layers. Greg Woodhouse 12:03, 12 May 2007 (CDT)
combined and renamed
I saw no reason for two short articles rather than one longer one, so I combined these under the name I am most familiar with.Pat Palmer 21:36, 12 May 2007 (CDT)
can we move the standards justification stuff?
I think the section justifying the need for network standards belongs on the computer network page, rather than here.Pat Palmer 21:36, 12 May 2007 (CDT)
- I think that seems sensible. Maybe something along the lines of "for further discussion, see Computer network" would be helpful. But, then again, maybe even that is not needed. Greg Woodhouse 22:25, 12 May 2007 (CDT)
- To get to basics, what is the perceived need here for "justifying" network standards? I could put more on Computer networking reference models, and indeed go before computer networking to such things as the Kingsbury Compromise of 1913, in which AT&T agreed to give up its telegraphy business in exchange for setting technical standards for nationwide telephony. (At this point, why didn't AT&T change its name to AT?)
- Anyway, the main driver for standards came from telephony and radio, well before computers. Once it was in the international arena, OSI was first driven by telephony, not computer people.
- Radio and telephony people know they must interconnect, so standards are important to them. With computers, however, most vendors stayed proprietary as long as possible, to "lock in" customers to their networks. The abortive attempt, in the 1980s, to force OSI protocols (as distinct from the reference models), failed for many reasons, not least of which is the then US-military-subsidized technical standards for the TCP/IP family were available for the cost of downloading electronic documents, while it cost thousands of dollars to get a set of the hard copy of OSI standards. Vendors often bundled TCP/IP free with UNIX systems, where an OSI stack (e.g., X.400 electronic mail) might cost $200,000.
- In other words, standards per se didn't justify OSI. OSI was a consequence of an environment that was accustomed to standards. That environment, however, could not compete economically with the TCP/IP protocol documentation and reference implementations. Howard C. Berkowitz 04:52, 7 May 2008 (CDT)
Internet protocol coercion
Let me start with the disclaimer that I do not believe anyone should be teaching or learning the original OSI 7-layer reference model, defined solely by ISO document 7498, because that definition is obsolete even to ISO. Quite a number of extensions were added, such as the OSI Management Framework, the OSI Routeing [sic] Framework, Internal organization of the network layer, etc.
I suppose we have to have the article, because so much information and misinformation refers to it. If I must use a layered model, which is not always the right tool, I use the 4-layer Internet model.
Nevertheless, I see something very common here: there is an attempt being made to show where Internet protocols fit into this model, and even what physical devices exist at various layers. There are two basic problems with doing so: the Internet Engineering Task Force consciously and explicitly does its protocol development without any concern for OSI compatibility, and the OSI reference model is an abstraction that defines no physical devices. RFC1812, Requirements for IP routers, is a pragmatic IETF definition, of a type ISO avoided "not to constrain implementers".
I have started network reference models to explain the relationships of the IETF, OSI, ATM, IEEE and other models, and how there are some decided incompatibilities. How can we avoid, in Citizendium, the perpetuation of OSI myths?
As an aside, I started working on the pre-ISO/ITU work in 1979, with the ANSI DISY project. I was the first engineer hired by the Corporation for Open Systems (COS), an industry group chartered to accelerate the introduction of OSI and ISDN protocols. In retrospect, I can look back at six years of my life, in which I was constantly preaching that "OSI is the answer", and see the fundamental problem: we may have had an answer, but we never had a useful definition of the question. Howard C. Berkowitz 11:14, 6 May 2008 (CDT)