OSI Model

Instead of "model" a better word would be to call it an architecture for communications between objects. It was open ended. So was the RFC process developed IP, for that matter.

Yep. In the 80s when this was being done it was very apparent that the folks running the ISO process were from the mainframe sized side of the camp. Of course this was when minicomputers and PCs were exploding and a single meg of ram was a big deal. For many systems at the time to give up all their ram to a protocol stack that no one was really using doomed it from the start. And to be honest TCP/IP didn't even work very well until the ram situation improved. These protocols need buffers to work well. :)

Worse. ISO was the international side of Ma Bell and the rest of the telecom industry which for many countries was part of it's national postal service.

The mainframe vendors were busy building their own proprietary networks; SNA (IBM) and DECNET (DEC) were the big ones. IP was in the universities and research labs. At least these products worked, within the family. It's a stretch to say that overall the X. standards ever came up to expectations.

Of course this was when minicomputers and PCs were exploding and a

Please check the info and link I posted in another message.

I think TinyTCP will run nicely in 50 kB. The bigger problem was there usually was no-one to talk to. When there was, no-one minded 50 kB of lock-down after around 1980 or so when at least 16 kb chips existed. Prior to that, UUCP ruled because it needed much less (2 kB?).

-- Robert

You'll use it to understand some new service you haven't heard of yet or to explain something to someone else. For instance:

I did, and didn't agree with it. Maybe we were quibbling over words. G*d knows enough words were spilled in this topic in the late

People forget that the full phrase was "ISO Reference Model". Google for that and you'll find article after article saying that the purpose was (and still is) to be a framework for interoperability.

For instance:

ISO in association with CCITT have produced a model for Open Systems Interconnection. This is a very loose standard which promotes the development of protocols designed to permit open systems interconnection. It also functions as a framework into which existing standards may be slotted. It should act as an aid in designing future protocol standards.

See this URL for an example of how the model is used to describe SS7, which is NOT compatible with the X.dot standards.

Which is fine...but it still is useless to me otherwise. Ok...so I know how a packet is created on one end and deconstructed on the other....phffft.

It currently is a "model", but as someone else mentioned, it wasn't a model for TCP/IP, which was in service before there was an ISO stack. You'll also find TCP/IP doesn't quite line up with ISO. Also, you can find a lot of stuff on the net, that isn't entirely accurate.

What do you mean by a "real example"? Its utility is that it provides a framework, not that you will be digging into the depths of it every day. If you don't know the OSI model and someone says "This is a Layer 3 switch" you won't have a clue what they're talking about. If you do know the model then you will know immediately that it has routing capability. To take one example.

How simple can we make this; In your career, if you have one, in computers and telecom you will routinely be presented with information structured with in a conceptual model described by ISORM. It ain't rocket science or magic. it encourages apples-apples comparisions which make learning faster and less error prone.

As for real products and systems, Any product that says it implements an X.something standard is something out of the ISO and to some degree is part of OSI. There are lots of X roducts in major use. Go to any tech library and look in _Newton's Telecom Dictionary_ under X.... and you'll find several pages of X.nnn standards.

Quick technical correction -- X. series standards (note the "dot") are from ITU-T (previously known as CCITT). The International Organization for Standardization (ISO) is a different beast, although today many of the ISO standards are also published as ITU-T standards. ISO standards can be recognized by the title "ISO/IEC-xxxxx" where xxxx is the standard number and ISO/IEC stands for "International Organization for Standardization/International Electrotechnical Commission". You may also see older ISO standards with an IS, DIS or DP prefix (for International Standard, Draft IS or Draft Proposal, respectively).

Note also that the IEEE and ANSI LAN and MAN standards are also under the ISO umbrella, and may get published as IS standards as they mature. For example, the stable IEEE 802.3 is also available as ISO/IEC-8802-3. Virtually every physical layer standard currently in use in the TCP/IP protocol architecture is an ISO/OSI product.

Technical trivia: ISO is not an acronym, it is the Greek prefix "iso" meaining "same," a pun on the goal of international standards. Note that ISO does far more than just networking standards. Long before computers were invented they were standardizing everything from fire hose couplings to screw threads. CCITT (which mutated to ITU-T circa 1990) was created by the League of Nations to standardize international telecommunications, they are the reason you have been able take for granted the ability to talk by phone or send a FAX worldwide.

The IETF and IAB are still learning how to deal with the impact of politics on technical standards as the monetary stakes involved rise in significance. Funny how as they do, they start to look more and more like the standards organizations they despised back in the 80's.

Getting back to the topic at hand... why learn the ISO OSIRM? There is no reason to "learn" it if your only goal is to get a job in networking. The world is full of network techs with a faulty understanding of OSI and network layering who are making a good living. Many of them even do a good job maintaining and even designing networks. Indeed, a thorough understanding of the OSIRM can be dangerous, as it can make it difficult to avoid disrupting meetings with fits of laughter as other's massacre the concepts and use flawed logic to make their points.

However, if your goal is to understand how networks work (and sometimes more important, how they can fail), a thorough understanding of the OSIRM is indispensible. As I tell students, networking is fundamentally simple, the confusion comes from the profusion of simple concepts which must be combined to create a functional system (and from bogus explanations of simple concepts by those who don't really understand them, and from the need for backwards compatibility with concepts and technologies which died out long before most practitioners were even born...we are still living with the remains of teletype support and the like).

If you want to be able to communicate clearly and concisely with other knowledgeable networkers, you need to be able to correctly use OSIRM terminology. And correct usage requires you to understand how layered protocols work and the standard terms used to describe those workings. Just as biology uses latin names to formally describe flora and fauna, so that confusion over local and colloquial names can be avoided (whether the same name is used for different species or different names used for the same species), network professionals need to be able to communicate fundamental networking concepts, and for better or for worse, the vendor & architecture neutral OSIRM is the language that has been selected for use. Your opinion of the relevance of the OSIRM to TCP/IP is irrelevant, even the most academic biologist will agree that Latin is a very dead language, but that does not make it any less useful for the purpose.

Unfortunately, the real world will measure your understanding not only by your ability to correctly use OSIRM terminology when communicating with others, but also your ability to recognize the misuse of OSIRM terminology by others and apply the breadth and depth of your networking knowledge to determine what concept they are actually trying to express.

Good luck and have fun.

Thats all nice to know info. But yet still NOBODY has any real example of how to apply this knowledge. Why is that?

There used to be a company called "Level One Communications," in Folsom, CA. They were so named because they made PHY devices. Level One was acquired by Intel a few years back.

A former member of the senior management of Level One left that company to form a new startup, primarily to make bridge/switch chips. Recognizing that the name of the company tended to restrict the business-thinking of company management (e.g., managers in a company called Level One might reject new, non-PHY product lines as out of the scope of the business model), he named his new company "Allayer Communications." That left him a lot of leeway. (Allayer was acquired by Broadcom.)

-- Rich Seifert Networks and Communications Consulting 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 (408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Whenever someone refers to the PHY interface on a mobo or other hardware the're refering to layer 1.

While that's true, you seldom see a PHY advertised as a "layer 1 interface". You do however see routers advertised as "layer 3 switches" and bridges advertised as "layer 2 switches", hence my use of that example.

Uh..yes, I've been working in IT for more than 6 years. And I know the OSI model, although the TCP/IP model seem more relevant, I have NEVER needed to count on that info for anything practical or NEVER has anyone provided anything as to when I would ever use it with the exception of being asked in a job interview.

And if you can't really answer my question, then don't talk down to me.

, in

Well, considering that the seven layers were a compromise between six and eight, one shouldn't take them too literally.

To me, the lower layers seem fairly obvious, and easy to keep apart. The higher layers I find much harder to understand the distinctions being made. Certainly for IP/ethernet the separation between two and three doesn't seem so hard, with IP being only one of the possibly ethernet protocols. I tend to believe that ICMP should be layer four, but many seem to believe layer three.

(snip)

-- glen

There are basically three reasons you need to know the OSI model, if you're going to work in networking:

1. You need to know the parlance of the field. Network Engineers commonly talk about things in terms of OSI layers - it's just how we talk. Engineers will talk about equipment in terms of layers. i.e. "Yeah, it's a layer 3 device, but it does layer 4 filtering." If you don't know what that means... well, then you don't know what that means.

1. It helps to understand a lot about data communications, in general, and the abstraction that takes place between layers. In other words, I don't care how you do it, as long as it accepts these inputs and provides those outputs. Each layer is built on that idea.

2. You will look really stupid in a job interview for a networking position, if the interviewer asks you to explain what you know about the OSI model and you scratch your head and say, "Uh, I know it has 7 layers..." And don't think they won't ask. I've done and been on plenty of technical interviews, and it usually comes up.

Just my thoughts...

grep

CJ wrote: