Talk:Computer networking reference models

To do. Please write a brief definition or description of this article at Computer networking reference models/Definition.

Metadata here

	NOTICE, please do not remove from top of page.
	I released this article to Wikipedia. In particular, the identical text that appears there is of my sole authorship. Therefore, no credit for Wikipedia content on the Citizendium applies.
	Check the history of edits to see who inserted this notice.

All materials taken from Wikipedia were in my private userpage, not Wikipedia mainspace, and contain only my words.

This is a work in progress, which will have introductions to the IETF, OSI, IEEE, and ATM reference models, each of which deserve its own full article.

Howard C. Berkowitz 04:42, 3 May 2008 (CDT)

IETF and the IAB have never intended to standardize a reference model

Mike Padlipsky is an entertaining and insightful writer, but RFC871 is a personal opinion. Having worked on OSI standardization (and, for that matter, the predecessor DISY work in ANSI in the late seventies) before I got involved in the IETF, the consensus in the IETF and Internet Architecture Board has always been that they do not want to be OSI-compliant and do not want a formal reference model. Where some of the techniques from OSI were useful, but they were adopted.

Requests for Comment fall into several categories, Standards Track (of three grades, Proposed Standard, Draft Standard, and Standard), Informational, Experimental, and Historic (i.e., obsolete and deprecated), although some of the earlier ones, such as Padlipsky's, are not characterized:

0871 Perspective on the ARPANET reference model. M.A. Padlipsky.

    September 1982. (Format: TXT=74455 bytes) (Status: UNKNOWN)

To the extent that layering has been used, but certainly not as a formal reference, the reference would be

1122 Requirements for Internet Hosts - Communication Layers. R.

    Braden, Ed.. October 1989. (Format: TXT=295992 bytes) (Updated by
    RFC1349, RFC4379) (Also STD0003) (Status: STANDARD)

Note that the above document was issued in 1989. The OSI Reference Model, ISO Document 7498, was published in 1984. If the IETF had intended to be compliant with OSI or its philosophy, they had the formal OSI specification available — and chose not to use it.

Nevertheless, there have been recent interpretations, often on behalf of the Internet Architecture Board, such as

3439 Some Internet Architectural Guidelines and Philosophy. R. Bush,

    D. Meyer. December 2002. (Format: TXT=70333 bytes) (Updates RFC1958)
    (Status: INFORMATIONAL)

RFC3439 contains the language, "Layering Considered Harmful": Emphasizing layering as the key driver of architecture is not a feature of the TCP/IP model, but rather of OSI. Much confusion comes from attempts to force OSI-like layering onto an architecture that minimizes their use."

STANDARD is a hard status to get. Very detailed specifications, which went through peer review in the Working Group and Internet Engineering Steering Group, still often come out as Informational. Been there, done that, have the T-shirt.

2072 Router Renumbering Guide. H. Berkowitz. January 1997. (Format:

    TXT=110591 bytes) (Updated by RFC4192) (Status: INFORMATIONAL)

4098 Terminology for Benchmarking BGP Device Convergence in the

    Control Plane. H. Berkowitz, E. Davies, Ed., S. Hares, P.
    Krishnaswamy, M. Lepp. June 2005. (Format: TXT=66845 bytes) (Status:
    INFORMATIONAL)

To get a sense of how the IAB does regard models, as a tool and not a rigid framework, see

4101 Writing Protocol Models. E. Rescorla, IAB. June 2005. (Format:

    TXT=47287 bytes) (Status: INFORMATIONAL)

I highly recommend Padlipsky's book, The Elements of Networking Style, as a hysterically funny book that encourages the reader to think deeply about networking. Nevertheless, I never had the impression that he was an advocate of rigid protocol reference models.

Even ISO considerably revised 7498, with four annexes, and additional documents such as "Internal Organization of the Network Layer", "OSI Routeing [sic]" Framework, and TR10000 on functional profiles. Seven layers went away long ago in everything except oversimplified educational material. Howard C. Berkowitz 23:15, 17 September 2008 (CDT)