Summary

  • RFC 1958, dated June 1996, prints "B. Carpenter, Editor" and the IAB in its header under Category: Informational, and its own status paragraph states that it "does not specify an Internet standard of any kind." (RFC 1958)
  • The category rule published four months later, RFC 2026 / BCP 9, limits what such a document may claim rather than what a reader may do with it; the 2002 successor RFC 3439 carries "Updates: 1958", stays Informational, and extends the same guidance. (RFC 2026)

The line above the principles

RFC 1958 runs to nine short sections. Before any of them comes the header block: Network Working Group, B. Carpenter, Editor, Request for Comments: 1958, IAB, Category: Informational, June 1996 (RFC 1958). Under it sits the Status of This Memo: "This memo provides information for the Internet community. This memo does not specify an Internet standard of any kind. Distribution of this memo is unlimited." Then the Abstract, whose second sentence most quotations leave behind: the principles are "intended for general guidance and general interest, and is in no way intended to be a formal or invariant reference model."

That status paragraph is the RFC Editor's boilerplate of the period, not a personal statement by the Editor named three lines above it. The distinction matters for anyone tempted to read the category line as modesty. It is not modesty. It is a series describing, before the reader reaches a single principle, what kind of object is in their hands.

What the principles actually ask for

Inside those covers the memo is direct. Section 1: "The principle of constant change is perhaps the only principle of the Internet that should survive indefinitely." Section 2.4: "Fortunately, nobody owns the Internet, there is no centralized control, and nobody can turn it off." Section 3.5: "Keep it simple. When in doubt during design, choose the simplest solution." Section 3.9: "Be strict when sending and tolerant when receiving." (RFC 1958) Section 4 is titled Name and address issues; 4.2 states "A single naming structure should be used", and 4.4 that addresses "must be unambiguous (unique within any scope where they may appear)". Section 5, External Issues, adds that an implementation missing required components "cannot claim conformance with the standard"; section 6.1 that "All designs must fit into the IP security architecture." (RFC 1958)

The word "must" appears repeatedly in a memo whose status says it specifies no standard of any kind. The tension is real, and neither sentence can be argued out of existence. The obligations are addressed to designs — what a design must do to fit. The category line is addressed to the document — what it may claim to be. A memo can bind a design decision through the force of its argument while declining, on its own first page, the authority of a standard.

The category rule that arrived four months later

RFC 2026, BCP 9, is dated October 1996, authored by S. Bradner of Harvard University, obsoletes RFC 1602, and carries Category Best Current Practice (RFC 2026). Section 4.2 places documents off the standards track under three labels — Experimental, Informational, Historic — and states that "The documents bearing these labels are not Internet Standards in any sense." Section 4.2.2 defines the Informational class as published "for the general information of the Internet community", saying it "does not represent an Internet community consensus or recommendation", subject only to editorial considerations and to verification that there has been adequate coordination with the standards process. Section 4.2.3 turns to misuse: the designations must not be "misused to circumvent the Internet Standards Process", and the IESG may insert disclaimer text where a memo conflicts with an established effort (RFC 2026).

Three provisions, three different addressees. The first defines status. The second defines what the class may claim about consensus. The third polices participants who would use a non-track label as a side door into the process. In the sections reviewed for this article, not one of them tells a reader, an operator or an implementer to disregard an Informational memo's guidance, and not one forbids adopting its advice. The limit is on the claim, not on the use.

One date deserves stating instead of smoothing. RFC 2026 is dated four months after RFC 1958. The reviewed text does not say which process rules governed the June 1996 classification, so the category line is a fact about the document as printed, not a verdict on the procedure that produced it.

A 1995 warning about precisely this

RFC 1796, "Not All RFCs are Standards", is dated April 1995 and signed by C. Huitema of INRIA, J. Postel of ISI and S. Crocker of CyberCash (RFC 1796). It calls it "a regrettably well spread misconception that publication as an RFC provides some level of recognition", and adds: "It does not, or at least not any more than the publication in a regular journal." It observes that a document's status "is sometimes omitted from quotes and references, which may feed the confusion", then predicts both halves of the market failure in two sentences: "Misguided vendors may claim conformance", and "misguided clients may actually believe that they are buying an Internet standard."

Written a year before RFC 1958, that warning describes what would happen to it.

The successor that stayed in the same class

The best evidence of what this category can carry is the document that came next. RFC 3439, "Some Internet Architectural Guidelines and Philosophy", is dated December 2002, authored by R. Bush and D. Meyer, carries "Updates: 1958" in its header, and is itself Category Informational (RFC 3439). Its status paragraph repeats the formula: "It does not specify an Internet standard of any kind." Its abstract says what it is doing — it "extends RFC 1958 by outlining some of the philosophical guidelines to which architects and designers of Internet backbone networks should adhere."

Six years after the first memo, the way the community added to the doctrine was to update it inside the same class. The 2002 text is no restatement. Section 2 makes simplicity the governing principle, holding that "complexity is the primary mechanism which impedes efficient scaling". Section 2.1 restates the end-to-end argument and then qualifies it: "the End-to-End Argument does not imply that the core of the Internet will not contain and maintain state" (RFC 3439). Sections 3 and 4 carry the titles "Layering Considered Harmful" and "Avoid the Universal Interworking Function"; section 9 says the document "attempts to codify long-understood Internet architectural principles" (RFC 3439).

Read side by side, the later text tightens an element of the earlier one: a reader who took the 1996 end-to-end principle as a statement that the core should hold nothing now has a 2002 sentence saying the argument does not imply that. That comparison is this article's reading of two printed sentences, not a claim RFC 3439 makes about itself. What RFC 3439 prints is "Updates: 1958" and "extends RFC 1958".

The force it actually carries

So what force does RFC 1958 carry? Not the force of a standard — its own first page says otherwise, and the category rule says it in stronger words. Not the force of an advisory that can be waved away either, because it has been extended on the record, by the same publisher, in the same class, by two named authors who put "Updates: 1958" where every reader can see it. The guidance travels by adoption. Engineers use it because it describes how the network behaves, and the record shows the community amending it by writing another memo rather than by elevating it.

What a reader can check is smaller and firmer than the reputation: the printed status line, the printed section numbers, and a 2002 header saying a later document updates the earlier one. Those three things are in the files. Whether an operator ever followed the simplicity principle is in none of them, and nothing reviewed here quantifies how widely the 1996 principles were implemented in operational networks.

A note on the form this reading rests on. Three documents were consulted in more than one rendering. The plain-text capture of RFC 1958 ends mid-section 3.8, so sections 4, 5 and 6 were read from the RFC Editor's HTML rendering; the plain-text capture of RFC 2026 ends before section 4.2.2, which was read from the RFC Editor's information page for RFC 2026; the plain-text capture of RFC 3439 breaks during section 2.2.1, and sections 3, 4 and 9 come from its HTML rendering (RFC 1958, RFC 2026, RFC 3439). The section number is printed beside every quotation above so a reader can reach the same place.