LUNAROPS · OPERATIONAL UPLINK 100% UPTIME 1,247d POSTS 893 JEFF.MOON@LUNAROPS.DEV UTC --:--:--

How Standards Bodies Actually Work

16 min read Law & Standards
standardsietfieeeprotocolsinteroperabilityfrand
Contents

Every time you open a socket, join a Wi-Fi network, or plug in a cable, you are trusting a document written by a committee you have never met, under rules you have never read, funded by companies whose interests do not align with yours. That this works at all is one of the quiet miracles of engineering. That it works unevenly — beautifully for email, miserably for USB naming, expensively for cellular — is the part worth understanding, because the quality of a standard is downstream of the institution that produced it, and those institutions could not be more different from one another.

There is no single “standards body.” There is a loose federation of organizations with incompatible philosophies about who gets to decide, how disputes resolve, whether the output is free, and what role patents play. The IETF runs on rough consensus and refuses to vote. The IEEE runs on formal balloting and corporate membership. ISO sells its documents for hundreds of euros a copy. The WHATWG is functionally four browser vendors in a trench coat. Knowing which body owns a standard tells you, before you read a single line, roughly how political it was, how openly it was developed, and whether you can even read it without a purchase order. This is a field guide to the major ones and how to actually participate.

The IETF: rough consensus and running code

The Internet Engineering Task Force produced TCP, IP, DNS, HTTP, TLS, BGP, and most of the plumbing under everything you do. It is also, structurally, the strangest of the major bodies — and the most quietly effective.

The IETF has no formal membership. You do not join; you show up. There are no dues, no corporate seats, no credentials. Participation happens on mailing lists and in working groups, and the standing rule, coined by David Clark in 1992, is the closest thing the internet has to a constitution: “We reject kings, presidents, and voting. We believe in rough consensus and running code.”

Both halves matter. Rough consensus is not unanimity and not majority rule — it is the sense of the working group that the remaining objections have been heard and addressed, even if not everyone is happy. A chair gauges consensus by humming in the room rather than counting hands, deliberately avoiding the optics of a vote, because a vote creates winners and losers and a faction that feels outvoted will not implement faithfully. Running code is the tiebreaker: if two designs are argued to a standstill, the one with two interoperating implementations wins, because the IETF’s entire reason for existing is interoperability, not theoretical elegance.

The output is the RFC — Request for Comments, a name that has been a polite fiction since the 1970s, because a Standards-Track RFC is anything but a request. The path runs from an Internet-Draft (the draft- documents, which expire after six months to prevent zombie proposals) through working-group adoption, IETF-wide last call, and approval by the IESG. RFCs are free, forever, readable by anyone, and that openness is not incidental — it is the reason the internet’s core protocols out-competed every closed alternative. You can read RFC 9110 (HTTP semantics) right now without asking permission or paying anyone.

The failure mode of the IETF is the inverse of its strength: it is slow, it favors incumbents who can afford to send people to three meetings a year, and “rough consensus” can be captured by whoever is loudest and most persistent on a mailing list. But the protocols that route your packets and the routing politics underneath them (BGP for engineers) came out of this process, and they have scaled across five orders of magnitude of growth. Slow and open beat fast and closed.

IEEE 802: where Wi-Fi versions are political documents

The IEEE Standards Association operates on the opposite model, and it owns the physical and link layers you touch most: Ethernet (802.3), Wi-Fi (802.11), and Bluetooth’s neighbor technologies live here. Where the IETF refuses to vote, the IEEE runs on formal, weighted, corporate balloting — and the difference in feel is enormous.

IEEE 802 working groups have members, and membership is earned by attendance and maintained by participation. Decisions go to ballot, often requiring 75% approval to advance, and a single sustained “no” vote with a technical comment must be formally resolved before the draft moves. This produces rigorous, heavily-reviewed documents. It also produces standards that are explicitly political artifacts, because the voters are engineers paid by Qualcomm, Broadcom, Intel, Cisco, and Apple, and each company arrives with a roadmap it wants the standard to bless.

Wi-Fi is the clearest example. The generation names you know — Wi-Fi 6, Wi-Fi 7 — are marketing labels the Wi-Fi Alliance bolted on later; the real documents are 802.11ax and 802.11be. Wi-Fi 7 (802.11be, “Extremely High Throughput”) brings 320 MHz channels, 4096-QAM, and Multi-Link Operation for headline rates up to 46 Gbps. Each of those features represents a negotiated settlement between companies that had already taped out silicon betting on a particular outcome. When you see a Wi-Fi spec take years, it is usually not because the physics is hard but because two chipmakers with incompatible implementations are fighting to have their approach become the mandatory one, since the loser eats a respin.

Here is the rough shape of an IEEE 802 standard’s journey from idea to product:

  Study Group ──> Task Group formed (e.g. 802.11be)
       │                  │
       │            draft + comment cycles
       │                  │  (each "no" vote needs resolution)
       ▼                  ▼
  Working Group ballot (>=75% approval)
       │
       ▼
  Sponsor ballot (wider IEEE-SA review)
       │
       ▼
  IEEE-SA Standards Board approval ──> published standard
       │
       ▼
  Wi-Fi Alliance certification + marketing name ("Wi-Fi 7")

The IEEE’s documents are also not uniformly free — many are available, but the association’s broader catalog includes paywalled standards, which brings us to the body that turned paywalling into a business model.

ISO/IEC and the pay-to-read problem

The International Organization for Standardization and its electrotechnical sibling the IEC sit at the top of the formal pyramid, the bodies governments recognize and treaties reference. They own everything from the C and C++ language standards to country codes (ISO 3166), date formats (ISO 8601), and the OSI model that every networking course still teaches. They operate through national body membership — ANSI for the US, BSI for the UK, DIN for Germany — so you do not participate as an individual; you participate through your country’s delegation, which is a meaningfully higher barrier to entry.

ISO’s defining and most criticized trait is that the documents cost money. The C++ standard, ISO/IEC 14882, is a several-hundred-franc purchase. ISO 8601, the date format every system on Earth uses, is paywalled — you can implement 2026-06-11 correctly only because someone summarized the spec for free elsewhere. This creates a genuine engineering problem: the canonical text of standards that underpin global interoperability is locked behind a cashier, so the working knowledge most developers have is secondhand, derived from blog posts and Stack Overflow rather than the normative source.

The pay-to-read model exists because ISO funds itself partly through document sales rather than entirely through membership and government grants, and defenders argue it keeps the organization independent. Critics — including many of the engineers who write the standards — point out that it directly undermines the goal of a standard, which is universal adoption. The C++ committee famously works around its own publisher by making the final working drafts freely available, so the document everyone actually reads is the free draft a few revisions behind the paid “real” one. When the people writing the standard route around the paywall, the paywall has lost the argument.

Body Membership Decision rule Documents Feel
IETF Open, no dues Rough consensus + running code Free (RFCs) Engineer-driven, slow, open
IEEE 802 Individual, attendance-earned Formal ballot, ~75% Mixed (some paywalled) Corporate, rigorous, political
ISO/IEC National bodies only Formal ballot, national votes Paywalled Governmental, formal, distant
W3C Paid org membership Consensus + Director (historically) Free Web platform, process-heavy
WHATWG Browser vendors (steering) Editor + implementer consensus Free, “living” Fast, vendor-controlled
USB-IF Paid corporate Promoter-group consensus Free spec, paid compliance Industry consortium

W3C versus WHATWG: the standard that forked and the fork won

The web has the most instructive governance story of all, because it is the one where a standards body lost control of its own flagship specification to a breakaway group — and the breakaway was right.

The World Wide Web Consortium, founded by Tim Berners-Lee, spent the early 2000s steering the web toward XHTML and a clean, XML-based future that the people building browsers did not want. So in 2004, engineers from Apple, Mozilla, and Opera formed the WHATWG — the Web Hypertext Application Technology Working Group — to develop HTML the way they actually intended to ship it. For years the web had two competing HTML specifications: the W3C’s versioned snapshots (HTML5, HTML 5.1) and the WHATWG’s continuously-updated “Living Standard” that never freezes and never gets a version number.

This is a textbook standards fork, and it resolved the way forks resolve when one side has the implementations: in 2019 the W3C and WHATWG signed a memorandum making the WHATWG Living Standard the single authoritative HTML and DOM specification. The W3C kept a role in process and review, but the browser vendors won, because they were the running code. A standard for HTML that the browsers do not implement is fan fiction. The lesson is the same one the IETF encodes deliberately: whoever controls the implementations controls the standard, regardless of which body holds the official letterhead.

The WHATWG model — a living standard, edited by people from the four engines that matter, updated continuously — is fast and pragmatic and has served the web well. It is also openly oligarchic: there is no pretense that an individual outsider has equal say with a Google or Apple engineer, because the steering group is the browser vendors. That honesty is arguably healthier than a body that claims openness it cannot deliver, but it is not democracy, and anyone proposing a web feature is negotiating with a handful of companies, not a community.

USB-IF and the art of the naming disaster

If the web shows standards governance at its most consequential, USB shows it at its most absurd — a cautionary tale about what happens when a consortium optimizes for marketing flexibility over user comprehension.

The USB Implementers Forum produces genuinely excellent technical standards. The connector is reversible, the power delivery spec pushes 240 watts, and USB4 rides on Thunderbolt’s PHY to hit real bandwidth. The names, however, are a catastrophe that has become a running joke in the industry. USB 3.0 was retroactively renamed USB 3.1 Gen 1, then renamed again to USB 3.2 Gen 1, all describing the same 5 Gbps link. The 10 Gbps tier became USB 3.1 Gen 2, then USB 3.2 Gen 2, then USB 3.2 Gen 2x1. A cable’s actual capability became unknowable from its label, which is the exact opposite of what a standard is for.

The USB-IF eventually admitted the problem and, with USB4 Version 2.0, moved to speed-based branding: the consumer-facing name is now simply “USB 80Gbps,” with logos that state the bandwidth and power directly on the cable and port. This is the right fix — describe the capability, not the spec revision — but it arrived after a decade of confusion that taught a generation of users to distrust USB labeling entirely. The lesson for any standards body is brutal and simple: the interface to the user is itself a design surface, and a technically perfect standard with an incomprehensible name has failed at its job, because a standard that humans cannot reason about does not interoperate in the only place that matters, which is the user’s head.

The same instinct that produced clean public APIs (API design principles) — name things for what they do, not how they were built — is exactly what USB spent ten years failing to apply to its own brand.

How the EU forced USB-C, and what that says about regulation

For most of standards history, governments stayed out of the technical details and let industry converge. The EU’s common-charger mandate broke that pattern and is worth studying as a model of regulation as a forcing function when voluntary standardization stalls.

For over a decade, the EU asked the industry to voluntarily converge on a single phone charger, and the industry mostly did — except for Apple, whose Lightning connector kept a large fraction of the market on a proprietary port. Voluntary consensus could not move the last holdout, because the holdout had a profitable accessory ecosystem to protect. So in 2022 the EU stopped asking and legislated. As of 28 December 2024, phones, tablets, cameras, headphones, handheld consoles, and a long list of other devices sold in the EU must charge over USB-C; laptops follow on 28 April 2026. Apple switched the iPhone to USB-C, worldwide, rather than build EU-specific hardware.

This is the rare case where a government wrote a connector into law, and it reveals the limits of voluntary standardization. Industry bodies are excellent at converging when the participants’ interests align and terrible when one powerful member benefits from fragmentation. When that happens, the only body with enough leverage to force convergence is a regulator with market access to withhold — the same lever that makes chip export controls bite. The trade-off is real: legislators move slower than technology and risk freezing a standard in place (critics worry the mandate could ossify USB-C past its useful life), but when consensus genuinely deadlocks, regulation is the only escape.

Patent pools and the FRAND bargain

Underneath many standards — especially cellular and video codecs — sits the messiest layer of all: the patents. A modern standard like 5G or H.265 is covered by thousands of standard-essential patents (SEPs), patents you literally cannot avoid infringing if you implement the standard, because the patented technique is the standard. This creates a hold-up problem: if implementing the standard requires infringing your patent, you could demand any price, and the standard would be unusable.

The settlement is FRAND — Fair, Reasonable, And Non-Discriminatory licensing. To get a technology adopted into a standard, the patent holder commits to license its essential patents to anyone, on fair and non-discriminatory terms. Patent pools then bundle many holders’ SEPs into a single license at a single price, so an implementer pays one pool rather than negotiating with hundreds of companies individually. This is the only reason you can build a phone that speaks 5G (how cell networks work) without a decade of bilateral patent litigation first.

FRAND is a permanent battlefield because “fair and reasonable” is undefined. Is the royalty a percentage of the chip or of the whole phone? Qualcomm and Apple spent years and billions litigating questions exactly that shape. The dynamic connects directly to how patents work for engineers generally (software patents for working engineers): a SEP is enormously more valuable than an ordinary patent precisely because the standard guarantees infringement, which is why companies fight so hard to get their technique adopted — winning the standard turns a patent into a tax on the entire industry.

How an individual engineer actually participates

All of this can feel like a spectator sport for trillion-dollar companies, but the open bodies are genuinely open, and getting in is more achievable than most engineers assume.

  • IETF is the easiest door. Join a working-group mailing list — they are public — and start reading. Comment substantively. File an Internet-Draft if you have a real proposal; anyone can submit one. Meetings have remote participation, and the culture genuinely weights good arguments over titles. The barrier is persistence and technical credibility, not money or credentials.
  • W3C community groups are open to anyone and are where many web features incubate before reaching the formal process. You do not need paid membership to join a community group.
  • WHATWG development happens in the open on GitHub. You can file issues and pull requests against the HTML and DOM standards directly; whether they land depends on convincing implementers, but the conversation is public.
  • IEEE and ISO are harder for individuals — IEEE 802 requires attendance to earn voting rights and ISO works through national bodies — so the realistic path is through an employer that already has a seat, or through your national standards organization.

The meta-skill is the same everywhere: standards are made by the people who show up consistently and do the unglamorous work of writing text, resolving comments, and building reference implementations. Influence accrues to persistence, not seniority. Much like the compounding visibility of building in public, the engineer who reliably shows up to a working group for two years ends up shaping the standard far more than a famous name who drops in once.

Verdict

There is no such thing as “the standards process” — there is a federation of bodies with opposite philosophies, and the body that owns a standard tells you almost everything about its character before you read a word. The IETF’s rough-consensus-and-running-code produced free, openly-developed internet plumbing that out-competed every closed rival, and it remains the cleanest model: slow, but legitimate and universal. IEEE 802 trades openness for corporate rigor and gives you heavily-reviewed but explicitly political standards where the votes are cast by the chipmakers. ISO sits at the top of the formal pyramid and undermines itself with a paywall so counterproductive that the people writing the C++ standard route around their own publisher. The web’s history proves the deepest rule of the whole field — whoever controls the implementations controls the standard, which is why the WHATWG fork beat the W3C — and USB proves the corollary that a technically perfect standard with an incomprehensible name has still failed. When voluntary consensus deadlocks on a powerful holdout, only a regulator can force convergence, as the EU did with USB-C. And under the cellular and codec standards you depend on sits the FRAND bargain, a permanent fight over what “fair” means for patents you cannot avoid infringing. The practical takeaway is that these institutions are more open than they look: the protocols you use were written by people who showed up, argued in good faith, and shipped running code, and that door is still open to anyone willing to do the same unglamorous work.

Sources

Comments