Talk:Open Systems Interconnection
|This is the talk page for discussing improvements to the Open Systems Interconnection article.
This is not a forum for general discussion of the article's subject.
|WikiProject Computing / Networking||(Rated Start-class, Mid-importance)|
Serious neutrality issues
I have corrected the worst NPOV problem, the claim that there were no nonproprietary network technologies before OSI. I have not corrected the others, for example the ridiculously biased claim that the OSI model was "the most important advance in education about networking". Looking at some of this text, I am actually seriously concerned that it may have been copied from the introduction to one of those long propagandist tomes about OSI published by researchers disgruntled that the Internet had passed them by in the late 1980s -- but I would have no idea how to figure out which one; these books are thankfully no longer in print.
No Wikipedia article should ever contain a claim of the form "X is the most important Y" as far as I'm concerned -- it is basically impossible to make such a claim from an NPOV.
Tls 20:22, 10 December 2006 (UTC)
Still in use
CLNP is still widely used today in many telecommunications networks (Mainly the transmission area) around the world. Still the new systems delivered utilise this. —Preceding unsigned comment added by 18.104.22.168 (talk • contribs) 09:53, 4 May 2004
- I think you're right. I just looked at some old ISO Bulletin articles I happen to have lying around and the term used in the 1980s was Open Systems Interconnection. Does anyone around here know how to move pages? --Coolcaesar 3 July 2005 04:06 (UTC)
- Holy bat-packet! I always though it was "connect", but I've just checked too, and you're right, it is "connection", e.g. "ISO 7498:1984 Open Systems Interconnection - Basic Reference Model". My only question is whether the early (70's) documents used "connect" - this one is '84. Let me check and see (and I just got finished moving it to Open Systems Interconnect, too - grrr). Noel (talk) 00:20, 2 September 2005 (UTC)
- Well, I've checked everything I've got, and I can only find "connection". I don't have any of the really early stuff (the first OSI Reference Model was published in '77 by the CCITT), but everything I have says "connection". So, I'll move the page... Noel (talk) 02:19, 2 September 2005 (UTC)
QS: X400 vs SMTP
- For example, the definition for OSI's X.400 e-mail standards took up several large books, while the Internet e-mail (SMTP) definition took only a few dozen pages in RFC-821. It should be noted, however, that over time there have been numerous RFCs which extended the original SMTP definition, so that its complete documentation finally took up several large books as well.
That's a horrible compromise of some POV argument, sounds much like a sandbox argument to me. Either you have a point or you don't. You cannot say, we have this strong argument in favor of something, but, then again, it should be noted that it's not really an argument. --22.214.171.124 09:41, 14 September 2006 (UTC)
You seem to have taken this out of context. Due to the nature of the now extint disagreement over the OSI Protocol's acceptability, this cannot be examined with an eye for technology and the biases that come with being a technophile. Instead, it must be considered with a mind for business.
Consider this when deciding whether or not it is an argument. The OSI took quite some time (a time period which includes the devlopment of the reference model as well) to fully develop and be published. Of those years, it can easily be assumed that only a portion of them were eaten up by the creation of QS:X400. It was several large volumes at creation. SMTP may now be several volumes in its own right, but this appears to be an evolutionary process rather than a birth defect as it was with the OSI.
It is not an argument over which is better. It is a simple statement of fact. Companies producing the necessary software and hardware for interconnectivity did not want to completely overhaul their production lines or code to accommodate the sudden appearance of the Encyclopedia Britannica of network models. It would cost them a fortune, and there was no guarantee it would catch on in the world at large. That appears to be the point being made in this statement. It would take weeks if not months of study just to fully read through and understand these protocols if the person is starting from the ground up. What would have made SMTP more acceptable for production purposes is the fact that it started out small and easy. In working its way toward massive, the bits added on later could be incorporated a little at a time because it already had a solid foundation. The adjustments in production would be minor and therefore more cost effective.
We implemented some pieces of X.400 at Bell Labs in the late 1980s. The Red-Book description of X.400 was not that big, but when the Blue Book came out, it was a telephone book! I included a comment in the article, based on a report which was published then (but sorry, I don't recall where) about the failure of French and German X.400 messaging systems to communicate. One was using delimiter-terminated strings and another was using a count/data format. These were options, and you didn't have to do implement both, so the two system were incompatable. A visiting scientist from IBM told us that it took several CPU seconds to send one X.400 message, in their experimental system. That seemed almost unbelievable to us...how naive we were about bloatware in those days.
At that time, SMTP was also heavily criticized for its complexity and almost numberless security holes. When 4.2 BSD came out, the email server brought the system to its knees and took minutes to send a message to more than a dozen people. The email problem seemed to attract a certain type of eager but not very talented mind. The mathematician Jim Reeds coined the phrase "Mailer Science" to describe the characteristic careless and bloated design style.
There are also horror stories about TP4 performance. It was a very inefficient protocol compared to TCP. I don't know how to communicate these problems here. I'm not going to stick unreferenced anecdotes into an article, but there were serious problems with OSI. It didn't just die because companies thought it was too expensive. It died in part because it was badly designed by folks who were in way over their heads. DonPMitchell 02:24, 19 November 2006 (UTC)
There may have been horror stories about TP4, but these were due to the implementations. There was at least one implementation that had just the same performance as other transport protocols around at the time. However the Transport Protocol as a whole (i.e. TP0-4) was a nightmare because of compromises between CCITT and others. TP0 came out of the Teletext world, and the others had to fit in around that. TP4 was TCP warmed over to fit in with the overall "architecture" of the TP family and OSI in general.
Of course OSI as a whole suffered generally from all the problems of design by committee. TCP/IP (and the other protocols that go with it) did an admirable job of providing everything that OSI was intended to provide, and with a far more tractable development process. If the OSI effort served any purpose, it was providing a standards-based vision that stopped SNA from becoming the de facto standard. In 1983 this seemed almost inevitable and was the justification for several companies' substantial investment in OSI. Jh493 (talk) 03:32, 27 March 2008 (UTC)
Proposed merge with "OSI model"
I'd say "no" - this page discusses the overall OSI project, while OSI model discusses a specific item, the OSI networking model, that this page claims antedated the project. Guy Harris 19:17, 22 November 2006 (UTC)
I also agree, the articles should not be merged. The OSI project is not the same things as the OSI model, which has been used to discuss many different network protocol schemes. DonPMitchell 21:53, 22 November 2006 (UTC)
I do not agree also , there is no reason to merge two articles one about a standards body and one about a standard by that body--Mancini 14:31, 26 November 2006 (UTC)
I don't agree either. No merge! 126.96.36.199 16:15, 10 December 2006 (UTC)
The Final Chapter
The article lacks a concise statement as to when and how the OSI vendors ceased promoting this solution and adopted TCP/IP, or withdrew from the market. 21:33, 31 December 2006 (UTC)
- First-hand experts could maybe illuminate re this?--The world salutes the Rising Star...Try to be One 10:11, 18 January 2007 (UTC)
The collapse of the OSI project in 1996 severely damaged the reputation and legitimacy of the organizations involved, especially ISO. The worst part was that OSI's backers took too long to recognize and accommodate the dominance of the TCP/IP protocol suite. The financial damage done to Japan and Europe (where internet deployment was delayed by years) is difficult to estimate.
Is it the generic internet(working technology) or the public Internet? Explain about the critique of the OSI Model and its protocols —Preceding unsigned comment added by 188.8.131.52 (talk) 12:28, 28 January 2011 (UTC)
Could someone please replace the adjective Byzantine with a more widely understood word. I am afraid to do any changes myself because I do not know exactly what it means. Velle 18:17, 7 March 2007 (UTC)
- The adjective is fine. It is a commonly used term in standard written English. The literal meaning (denotation) is anything associated with the Byzantine Empire. The figurative meaning (connotation), which is probably the one intended here, is anything that is mazelike, confusing, or overly complex. --Coolcaesar 22:36, 7 March 2007 (UTC)
This article seems biased against the people who opposed OSI. Where is the information from THEIR point of view? Here is something I put together for the Internet articles talk page.
Around 1982 the International Organization for Standardization (ISO) developed its own internetworking referance model called Open Systems Interconnections (OSI).
It was designed to replace TCP/IP and there was pressure to change over. Many large businesses, including computer development companies, as well as Europian governments and latter the American Government officially embraced OSI over TCP/IP.
This was while it was still just a design! This made things harder for those wishing to develop TCP/IP compatable technologies, such as MCI Mail.
In Europian universities something of a TCP/IP movement established itself.
In 1988 they finally developed it.
For a while it was likely that OSI would become the standard internetworking.
However, OSI was new and what you could do with it was not known, where TCP/IP was proven fact, and was already well established.
One key developing that helped deside the matter between OSI and TCP/IP was the giving away of TCP/IP protocals in Berkley Unix software that came with computers.
Where Wizards Stay Up Late by Katie Hafner and Matthew Lyon.
- Please don't: the dates are wrong and the statements (for the most part) incorrect. -- Dgtsyb (talk) 04:14, 16 May 2008 (UTC)
I admit I may of misread that part of the book, and it was not something they into depth in. However, I think we should look at what the people who opposed OSI were thinking when they opposed it. Corrupt one (talk) 23:41, 16 May 2008 (UTC)
- Well, step one to doing that would be identifying (if there truly were) the people that "opposed" OSI. Who were those people? Any names? What were their roles, interest and affiliation at the time? I don't remember anyone "opposing" it. Can you point to a verifiable primary source of this claim that there were people "opposing" OSI? It just doesn't jive with what I recall from first-hand experience back in the day, as well as what can be gleaned from, say, the history captured by the IETF RFC database. The IETF has always run by concensus and the concensus in most RFCs published between about 1991 and 1995 is that OSI was going to replace TCP/IP (and did for a number of application areas, and has to this day for a number of application areas, such as aviation, military, telecommunications management, manufacturing automation, ...).
- Now, BSD'ers perhaps? Don't think so. The Berkeley Sockets interface was redone to support OSI, and OSI was provided in BSD4.3 Reno and BSD4.4 (according to McKusick, who should know what he was saying (see, "The Design and Implementation of the 4.4BSD Operating System", section 13.9 on page 478-479). Which also refutes the availability of the BSD implementation of the TCP/IP protocol stack shipping with BSD that made it popular, because the BSD implementations of OSI: CLNP, CLNS, TP4, CONS, X.25, also shipped with the self-same BSD releases (yes, TCP may have initially been in BSD 4.2, whereas OSI was not available until 4.3 Reno, but so what). The situation is likely the reverse: that because commercial UNIX did not have a usable OSI protocol component until SVR 4, and because OSI was never shipped with the base SVR 4 UNIX operating system (and was a, sometimes quite expensive, add on), it may very well be that it is the lack of a commercial implementation of OSI protocols shipped with the AT&T UNIX operating system that caused its lack of popularity. -- Dgtsyb (talk) 01:29 17 May 2008 (UTC).
I am not really into programing history, but I remember from "Where Wizards Stay Up Late" by Katie Hafner and Matthew Lyon, which was made with the help of the people who were involved with the creation of the Internet, that there was opposition. The area that deals with OSI is a small part in the chapter called "A rocket on our hands." It mentions opposition to OSI, and why they were against it.
I am interested in the RFCs. Where can I find the address for them?
- Whatever after-the-fact anecdotal information claiming credit for predicting the unexpected result of the use of OSI may appear in that book of rather questionable credibility, the record of the Internet community (RFC Database) shows the Internet community to be in support of OSI. The record does show strong and heavily biased opposition against the use of X.25 as the (only) Network protocol for OSI instead of (at least also) the Internet protocol (or some datagram based protocol as was later provided by CLNP) (see RFC 874), but even this rather caustic opposition to X.25 is in fact in support of OSI. -- Dgtsyb (talk) 05:13 18 May 2008 (UTC).
- I think it was Bernard Aboba, if I recall correctly, who took the lead in terms of bashing OSI and ISO in his books.
- Also, other major components of OSI had severe problems. For example, ASN.1 is notoriously difficult for laypersons to understand and fiendishly difficult to implement properly, yet ISO wanted to use ASN.1 as the primary encoding for Office Document Architecture, upon which OSI's hypertext system, HyperODA, was to be based. Not surprisingly, ODA and HyperODA were never developed to the point where there was actually a working product with widespread deployment. In contrast, almost anyone who could use a word processor (myself included) could do trial-and-error with drafting simple HTML Web pages and feeding them into Mosaic until the concepts of tags, elements, attributes, and the like became clear. --Coolcaesar (talk) 06:26, 19 May 2008 (UTC)
- I doubt that Katie Hafner and Matthew Lyon were talking about Bernard Aboba.
- Also, ODA and HyperODA, while ISO standards or ISO draft standards, are not OSI components: they are file formats and not communications protocols. --Dgtsyb (talk) 11:12 19 May 2008 (UTC)
- Actually, if you search on this topic on Google or Google Books for five minutes, it's pretty clear that ODA and HyperODA were always positioned from the beginning by the OSI community as application layer formats for OSI, much as HTML serves as an application layer today for TCP/IP. OSI wanted to standardize the entire protocol stack, kind of like how America Online not only made their own server and client software but even made their own proprietary multimedia document format, Rainman. The line between static document format and dynamic interface protocol has always been kind of hazy.
- Also, I wasn't saying that Hafner and Lyon were talking about Aboba; I was saying that it was Aboba, NOT Hafner or Lyon, who really took the lead in the mid-90s in publicly bashing ISO and OSI and glorifying IETF and the RFC process. --Coolcaesar (talk) 05:27, 20 May 2008 (UTC)
- Computer Networks by Tanenbaum also describes criticism. The standards were difficult to implement, implementations were of poor quality, and people started thinking of bureaucrats trying to shove down their throats a poor standard, instead of a vastly better one. He also compares this with governments pushing PL/I or Ada as "the language of the future". I'm sorry I can't provide exact citations (I have the Italian edition and I am "backtranslating" it), but just look at § 1.4.4 - it's available on Google books (look at page 48 for this one).
I've not read it, but it may be useful to read this paper, which is criticism about implementability, done back in 1984: http://portal.acm.org/citation.cfm?id=639624.802077&coll=ACM&dl=ACM. --Blaisorblade (talk) 02:07, 19 July 2008 (UTC)
After reading this subject within wiki I decided to write a view below... And being in the thick of it all... There are a few issues which I and many others certainly could never get to the bottom of..
1. OSI was called a protocol and still is being referred to that way. - when simply OSI isnt "a protocol" its a 7 layer distributed systems / communication reference model... X.500 suffered the same description too in that it was called a "complex protocol" when it is in fact a directory service standard that has naming and information object definitions and an access and distribution protocol. So in there lies an issue.. If we call something a protocol which isnt a protocol, we can get confused quite quickly.
2. OSI was seen as too many bytes on the wire.. CLNP had a longer address space than IPv4 as it could use E.164 phone numbers (IPv4 then then moved to IPv6 and do see RFC 1888) and TPT 4 and TCP were very similar - I wrote TP4 for mainframe stream based communications in 1983 in about 2 months. It was fast... Later down the track we had TCP services too.
3. OSI was also called complex... Many of us did not find it any more complex than SNA, Decnet, ICLs architecture or the Internet protocols. E.g In about 1995 I was told X.500 is so complex - I could only say you should see the wiring and electonics of my Jeep if you want to see "complex". We ran X.500 over TPT and TCP - no problems. IBM had TN 3270 which was SNA LU2 over TCP/IP - the point being protocols stacks were now evolving and converging on TCP/IP
4. The use of ASN1 for application level protocols and their presentation layer services - meant that if one had an ASN.1 compiler, one could take the electronic copy of the standard (X.400, 500, 700, etc) and feed the protocol definitions into it and out would pop all the C and .H files for all protocol encoding and decoding processes (the presentation layer). Of course implementing the application layer's protocols/services would be difficult if one did not have an ASN.1 compiler..(the right tools) I am not saying these services are easy to implement BTW, but using the using good tools does make it easier. ASN.1 is used in many Internet protocols such as OCSP LDAP and PKI services.
5. We also worked on SGML in the 1980s - which sort of merged with the work on HTML..
I think a lot of mis interpretation went on within this debate in the 80s and 90s .. And I never did understand why some messages got conveyed.. We first and foremost wanted to develop a global common reference model and within that fit profiles of protocols..and a common approach to naming and addressing in order to achieve a common language around distributed computers from differen vendors. I never saw why ISO / ITU protocols were called complex or slow - as engineers we made them as simple, reliable and fast as possible. SS7 seems to be doing all right still. I have never seen a comparison report either where like by like demonstrates a protocol engineering issue.
We cant fix history and we all have shared in its outcome.. All I am hoping is that now we can see and understand what was said at the time with the knowledge people had - and that some of the messages conveyed about OSI at the time were IMHO and engineering experience (and many, many others) very diffucult to understand.
Lack of description
In reading this article, there is an indepth look into the techonologies history but it lacks a decent explanation of what it actually is. If there is a person with indepth knowledge of the techonology who can expand on this it would further round out the article. Baellen (talk)Baellen —Preceding undated comment was added at 07:01, 11 August 2008 (UTC)
- My thoughts exactly. I'll stick it on my list of things to do. I can't really claim to be an expert on OSI but I'm reading a book dealing exclusively with it at the moment purely out of interest. OTOH I'm reluctant to spend too much time researching and referencing an out of date protocol. CrispMuncher (talk) 21:45, 18 September 2008 (UTC)
This "talk" page is full of mentions of CLNP, X.400, TP4, ASN.1 etc -- shouldn't the various standards and protocols that are thought of as "OSI" get at least a mention and link in the article itself? — Preceding unsigned comment added by 184.108.40.206 (talk) 03:32, 28 June 2011 (UTC)
Edits by 220.127.116.11
I reverted edits by 18.104.22.168 because they seem to be POV and don't incorporate well in the article. I am moving the section added by 22.214.171.124 for reference. Nuβiατεch Talk 09:31, 14 April 2009 (UTC)
Its 25 years on so hopefully another view may help balance the criticisms
Background I spent 16 years with ISO/ITU and some with the IETF.. I was engaged in ICL developing multi nodal mainframes back in 1983. ICL was one of the EUs Mainframe computer supplier and Jack Houldsworth within ICL was considered the "grandfather" of OSI. At that time each computer vendor had its own comms architecture and terms about protocols and service interfaces, addressing, session control and data language making even talking about distributed communications and interworking almost impossible. Hence the need for a common model - an OSI model.. 7 basic interchange and service layers where each layer can be assigned protocol types to that pertinant to that layer - and each layer can expose a service interface to the layer above - A service interface being very similar to Unix Sockets..
The "layer" and service interface language of OSI is widley used today. So point 1. The critical one - OSI is a 7 layer reference model- thats all.
Protocols can be related to OSI such as Link protocols, Network, Internetwork, Transport, Session and above.. Even using these basic words about layers means we can all understand where such protocols fit in the model.
At my time in ICL in 1983 - We developed a full OSI stack for communications between mainframes for inter CPU O/S communications, disc I/O and comms I/O running over 50mbit LANs - and achieved a TPT level turnaround in 11 microseconds. So slow and complex was never my understanding..
Point 2. The OSI protocol suite - I am not quite sure where that one came from - there were protocol profiles, yes, such as GOSIP, but the intent was to have distributed systems with a common communications architecture and naming and addressing schema in which standard protocols could be used. GOSIP did include TCP/IP at later stages. And please bear in mind - in these days each country, particularly the EU and commonwealth we were trying to normalise its IT systems from propretary networks ..We were working with what we had and knew at the time.
Point 3. Naming and Addressing - the essentials of distributed communications - Often overlooked if protocols are considered as just the means to transfer data. CLNP the EU version of IP for instance allowed NSAP address forms which could have used E.164 telephone numbers - meaning that PSTN/VoIP type services with TNs were possible with CLNP.
Point 4. OSI had to consider all forms of communications in order to develop a generic model and provide language for all forms of new and existing communications - as it certainly was not a green field.. All mainframe vendors had interests - OSI even had to consider G3 Fax machines which use ITU transport class 0 and session protocols - and we needed to include telcoms protocols of the day which used physical, link and network signalling.. for PSTN, ISDN, FRX - We even considered Decnet and SNA other vendor systems.
The age of Internetworking.. During the 1980s Internetworking started its life as did the need for common distributed networking language as they seemed to go hand in hand. Most of us in the "ISO/ITU" OSI world of the time were working with TCP/IP in some form as well as SNA (IBM), DEcnet, etc - so we welcomed the Internetworking evolution...and the eventual retirement of all those proprietary bits and bytes on the wire.
We then moved up the stack... For example X.400 was the EU carrier grade for messaging, X.700 was for management and X.500 for directories.. all by design used an ASN.1 compiler to automatically build its application and presentation layer protocol processes.
The "discussions" about OSI certainly did see X.400 get replaced by SMTP, and the merging of management protocols and service - re X.700 standards and SNMP. But X.500 a distributed directory service in which we defined information objects and naming constructs first and then access and distribution protocols last seemed to be labelled as a complex protcol for some reason - so LDAP was born - which is really X.511 the Directory Access Protocol in string encoded form. Our engineering experience - we devloped a "full blown" X.500 directory service with LDAP. X.511 DAP stack took 75kb of code space, LDAP three times as much. X.500 resided in about 50mbytes of code space.. The same as an office automation package of the time.We now have broadband and gigabytes - so such debates are now history.
Its a waste of time to discuss winners or loosers now. I like to think that all of us won as we are here today with many forms of communications and distributed systems which we can build and exchange views on.. The 1980s was an age where many had strong views, where many of us who were chairing the standards processes and liasing between them were involved with the battles but really wanted the situation to normalise. After all it was the age of global internetworking.
From my own perspective X.500 (and X.509) started me on the road of large scale directory services and identity engineering. In my view LDAP did two things - The down side was it managed to convey a completely wrong message that a distributed directory service with common naming properties and object definitions and security (eg. X.509 and certficates) was just a "protocol" - a complex protocol. This is like saying that SQL is also the database and its data sets. But the big upside is that LDAP put directories, both small and large scale systems and identity management (IDM) onto the world stage.. IDM has become a critical issue in all systems today.
25 years on we are now working on converged internet services, identity managed service delivery platforms..and next generation event based directory engines.. see www.wwite.com
- All the above may be true, but I did a lot of OSI based networking consultancy for a number of UK companies and government departments in 1986 and 1987, and at that time it was fairly obvious on the ground that OSI was a strategic device to benefit European hardware manufacturers like ICL and Bull and disadvantage American suppliers, especially IBM. I never met a single fellow techie at the time who thought OSI was a good choice on technical grounds. --126.96.36.199 (talk) 12:33, 5 August 2009 (UTC)
Neutrality of Criticism
I just read the article, including the section on 'Criticism'. While I agree pointing out criticism is a good thing, especially in a case like this where an attempt at standardising is being discussed, it seems rather one-sided. For one, only criticism is mentioned without mentioning the advantages of the attempt. I can't believe there's nothing good in this, especially since the 7-layer model still seems to be used widely. Shouldn't a section about advantages be included as well? (Or perhaps the section criticism should be changed to 'Advantages/Disadvantages'?)
My other point of concern in this section is that only one source is cited, while there are several aspects of criticism. If we're just shown one source, how can we trust this is widely accepted at all? Could someone provide more sources in order to increase credibility of the section?HSNie 07:57, 9 January 2015 (UTC) — Preceding unsigned comment added by HSNie (talk • contribs)