School of Specs How standards actually get madeIn depth

Cheat sheet

How standards actually get made

The working reference on 3GPP's and the IETF's machinery

Depth
In depth
Chapters
16
Reading
120 min
Questions
84
Document
TR 21.900 v20.0.0
Sources cited
146

Back to the course

The shape of it

1 Why any of this exists

  1. 1 Why anybody writes a standard down

2 How 3GPP is put together

  1. 2 The shape of 3GPP, and what the record does not say about it
  2. 3 What a Release is, and what freezing one costs

3 How a 3GPP document is made

  1. 4 The life of a 3GPP document, from blank page to published version
  2. 5 The Change Request, field by field
  3. 6 Work Items and Study Items — how the work is authorised
  4. 7 Reading a 3GPP number — the series, the version, the file name
  5. 8 How a 3GPP meeting decides, and the word it decides in
  6. 9 Why the documents read the way they do

4 How the IETF works

  1. 10 The shape of the IETF
  2. 11 The life of an RFC, from a draft nobody may cite to a number nobody may edit
  3. 12 Rough consensus, and why the room hums
  4. 13 MUST, SHOULD and MAY — three words, one short document

5 Reading the record yourself

  1. 14 The two cultures, side by side and on the evidence
  2. 15 Reading either body's paper trail from a cold start
  3. 16 What this record cannot tell you

Keep these in your head

Easy to get wrong

One from each chapter that has one. The course carries 26 in all.

  • 1.3 Why anybody writes a standard down

    It is tempting to read "not a legally constituted entity" as "not serious". The opposite is true: the partnership pushes legal weight outwards to the Organizational Partners, who are standards bodies in their own right and who republish the output as their own standards. Being available on the 3GPP file server is not the same act as being published §5.1.

  • 2.7 The shape of 3GPP, and what the record does not say about it

    "Meets most often" is not "does the most", and "owns the most specifications" is neither. The columns disagree on purpose. A sentence that picks one column and calls it the workload has picked the answer it wanted, and the other two columns are sitting right there saying otherwise.

  • 3.4 What a Release is, and what freezing one costs

    "Not propagated" is not the same as "withdrawn". A specification that stops at Release 18 still exists at Release 18 and is still readable; it just is not part of Release 19. Withdrawal is a separate, heavier act §4.9.

  • 4.5 The life of a 3GPP document, from blank page to published version

    The jump is not a version being skipped. The first digit stopped meaning maturity at that moment and started meaning Release. Reading 4.0.0 as "the fourth revision" is the single most common misreading of a 3GPP version number. Reading a 3GPP number — the series, the version, the file name sets out what each digit carries.

  • 5.5 The Change Request, field by field

    The boundary between F and C is where arguments happen, and it is worth knowing which way it cuts. F is available on a frozen Release; C is not. So calling a change a correction rather than a functional modification decides whether it can go in at all — which is why the categories are checked, not merely declared.

  • 6.3 Work Items and Study Items — how the work is authorised

    "Four Member Organizations" is not a formality. It is the point at which a company's private wish becomes a shared piece of work, and it is why proposals circulate for support before they are ever tabled. Nothing in the working methods says what happens if a supporter later withdraws.

  • 7.2 Reading a 3GPP number — the series, the version, the file name

    The series is not a ranking and the number is not a version. TS 23.502 is not a newer TS 23.501, and the 38-series is not "better" than the 36-series. Both halves of the number describe subject matter, nothing else.

  • 8.2 How a 3GPP meeting decides, and the word it decides in

    A liaison statement is a document, not a letter. It is how one group formally asks another a question or tells it something, and it leaves the same paper trail as any other contribution — which is why 47,474 of them are on record [1]. That is the third largest of the five types counted here, and still a small part of the 1,629,643 meeting documents on record [2].

  • 9.3 Why the documents read the way they do

    "Should" is not a polite "shall". A product that ignores a "should" still conforms. A product that ignores a "shall" does not. If somebody quotes a "should" at you as an obligation, they have misread the document, and the drafting rules are the place to say so.

  • 10.2 The shape of the IETF

    This course cannot tell you how many working groups or areas exist, or what they are called. RFC 2418 spoke of more than 100 working groups and 8 areas — in 1998. No file read here gives today's figures, and the RFC itself says the current list lives on the website because it changes [3]. Anybody quoting those numbers as current is quoting RFC 2418, published in 1998, not today's list.

  • 11.3 The life of an RFC, from a draft nobody may cite to a number nobody may edit

    "Last Call" sounds like a formality and is not. It is the moment at which the document leaves its working group's hands and is exposed to everyone who did not attend. A document that only ever convinced the people who wrote it tends to find that out here.

  • 12.3 Rough consensus, and why the room hums

    The commonest misreading is to treat rough consensus as "most people agreed". It is closer to "no unanswered technical objection remains". A single well-made objection that nobody can answer blocks the work; forty people who simply prefer the other option do not, on their own, carry it.

  • 13.3 MUST, SHOULD and MAY — three words, one short document

    Treating SHOULD as "optional" is the most expensive misreading in the series. Two products that each ignore a different SHOULD, each for a locally sensible reason, can fail to work together while both remain conforming. That is not a defect in either product; it is what a SHOULD means, and it is why arguments about whether a rule should be MUST or SHOULD take so long.

  • 14.4 The two cultures, side by side and on the evidence

    That last asymmetry is a gap in the evidence, not a finding about the two bodies. It would be easy to read "the IETF's rule is open and 3GPP's is unknown" as a verdict. It is not one. It means one document was available here and the other was not.

  • 15.3 Reading either body's paper trail from a cold start

    The 21-series is not all process documents. TS 21.111 sits in the same range and is a requirements specification about the USIM card TS 21.111 v19.0.0. "It's a 21-series document, so it is about working methods" is wrong often enough to catch people.

  • 16.7 What this record cannot tell you

    The most common way a course like this goes wrong is not inventing a fact. It is letting a true structural comparison slide into a verdict — noting that the IETF requires implementation evidence and 3GPP does not, and letting the reader conclude which is better. The evidence supports the first half of that sentence and says nothing about the second.

The numbers, and where each came from

Every one was counted out of a file on this machine. Hover a row to read the question that produced it.

  • 97 97 clauses in TR 21.900, the working-methods document
  • 579 579 terms defined in the 3GPP vocabulary
  • 3928 3,928 3GPP specifications in the catalogue
  • 2149 2,149 of them are Technical Specifications
  • 1778 1,778 of them are Technical Reports
  • 3599 3,599 specifications are marked still maintained
  • 329 329 specifications are marked no longer maintained
  • 9735 9,735 RFCs in the local archive
  • 4371 4,371 of them are 3GPP meetings, from 1998-12-07 to 2028-06-13
  • 1629643 1,629,643 meeting documents recorded
  • 516809 516,809 of them are change requests
  • 27 27 3GPP bodies hold a meeting series of their own
  • 331 331 3GPP meetings were held in 2024 and 2025
  • 86887 86,887 delegate registrations were recorded across them
  • 213886 213,886 documents were filed to them
  • 21 21 3GPP bodies held at least 4 meetings in 2024 and 2025
  • 15 3GPP RAN 4 held 15 meetings in 2024 and 2025
  • 541 541 delegates registered for the middle 3GPP RAN 4 meeting
  • 3079 3,079 documents were filed to the middle 3GPP RAN 4 meeting
  • 38154 38,154 documents were filed to 3GPP RAN 4 in 2024 and 2025
  • 16 3GPP SA 2 held 16 meetings in 2024 and 2025
  • 453 453 delegates registered for the middle 3GPP SA 2 meeting
  • 1694 1,694 documents were filed to the middle 3GPP SA 2 meeting
  • 24355 24,355 documents were filed to 3GPP SA 2 in 2024 and 2025
  • 13 3GPP RAN 2 held 13 meetings in 2024 and 2025
  • 702 702 delegates registered for the middle 3GPP RAN 2 meeting
  • 1617 1,617 documents were filed to the middle 3GPP RAN 2 meeting
  • 19478 19,478 documents were filed to 3GPP RAN 2 in 2024 and 2025
  • 14 3GPP RAN 1 held 14 meetings in 2024 and 2025
  • 1092 1,092 delegates registered for the middle 3GPP RAN 1 meeting
  • 1600 1,600 documents were filed to the middle 3GPP RAN 1 meeting
  • 20556 20,556 documents were filed to 3GPP RAN 1 in 2024 and 2025
  • 13 3GPP RAN 3 held 13 meetings in 2024 and 2025
  • 425 425 delegates registered for the middle 3GPP RAN 3 meeting
  • 938 938 documents were filed to the middle 3GPP RAN 3 meeting
  • 11384 11,384 documents were filed to 3GPP RAN 3 in 2024 and 2025
  • 17 3GPP SA 5 held 17 meetings in 2024 and 2025
  • 203 203 delegates registered for the middle 3GPP SA 5 meeting
  • 912 912 documents were filed to the middle 3GPP SA 5 meeting
  • 12437 12,437 documents were filed to 3GPP SA 5 in 2024 and 2025
  • 13 3GPP CT 1 held 13 meetings in 2024 and 2025
  • 231 231 delegates registered for the middle 3GPP CT 1 meeting
  • 907 907 documents were filed to the middle 3GPP CT 1 meeting
  • 12100 12,100 documents were filed to 3GPP CT 1 in 2024 and 2025
  • 19 3GPP RAN 5 held 19 meetings in 2024 and 2025
  • 28 28 delegates registered for the middle 3GPP RAN 5 meeting
  • 289 289 documents were filed to the middle 3GPP RAN 5 meeting
  • 15199 15,199 documents were filed to 3GPP RAN 5 in 2024 and 2025
  • 29 3GPP SA 3 held 29 meetings in 2024 and 2025
  • 61 61 delegates registered for the middle 3GPP SA 3 meeting
  • 145 145 documents were filed to the middle 3GPP SA 3 meeting
  • 10375 10,375 documents were filed to 3GPP SA 3 in 2024 and 2025
  • 70 3GPP SA 4 held 70 meetings in 2024 and 2025
  • 32 32 delegates registered for the middle 3GPP SA 4 meeting
  • 16 16 documents were filed to the middle 3GPP SA 4 meeting
  • 5588 5,588 documents were filed to 3GPP SA 4 in 2024 and 2025
  • 53714 53,714 of those registrations ended with a presence recorded
  • 685 685 specifications carry the group code S5
  • 458 458 specifications carry the group code R4
  • 343 343 specifications carry the group code S3
  • 321 321 specifications carry the group code S2
  • 303 303 specifications carry the group code S4
  • 256 256 specifications carry the group code S1
  • 237 237 specifications carry the group code C1
  • 236 236 specifications carry the group code C4
  • 200 200 specifications carry the group code RP
  • 19 19 named releases
  • 20793 20,793 specification-and-release pairs
  • 97521 97,521 meeting documents ended with the status "approved"
  • 16137 16,137 meeting documents ended with the status "not pursued"
  • 159055 159,055 of them are pseudo change requests, for documents not yet under change control
  • 11595 11,595 of them are new work item descriptions
  • 5532 5,532 of them are new study item descriptions
  • 47474 47,474 of them are liaison statements sent to another group
  • 1543066 1,543,066 of them carry an upload date, running from 2015 to 2026
  • 8457 8,457 of them are plain text, numbered 1 to 8649
  • 1278 1,278 of them are XML, numbered 8650 to 9953

The document's own words

One quotation per chapter that carries one, of 20.

Where to look it up

The 46 clauses of TR 21.900 this course is built from.

§Foreword§Introduction§1§1A§2§3§3.1§3.2§4.0§4.0A§4.0B§4.1§4.1.1§4.1.2§4.2§4.3§4.4§4.5§4.6§4.6.5§4.6.6§4.7§4.8§4.9§4.9A§4.9B§4.10§4.10.2§4.10.3.4§5.1§5.2§5A§5B§6.0§6.1§6.2§6.3§6.4§6.5§6.6§6.7§7§7.5§8§9§Annex A

The 69 named sources it cites, by what they are.

Working group 21

3GPP OP3GPP SA3GPP RAN3GPP CT3GPP PCG3GPP3GPP SA 13GPP RAN 13GPP CT 13GPP SA 23GPP RAN 23GPP CT 33GPP SA 33GPP RAN 33GPP CT 43GPP SA 43GPP RAN 43GPP CT 63GPP SA 53GPP RAN 53GPP SA 6

Quoted wording 17

RFC 3935, section 1TR 21.801, Annex ETR 21.801, Annex ERFC 2418, section 1RFC 2418, section 1RFC 2026, section 2.2RFC 2026, section 4.1.1RFC 6410, section 2RFC 2026, section 1.1RFC 7282, section 1RFC 7282, section 1RFC 2418, section 3.3RFC 2418, section 3.3RFC 2119, section 1RFC 2119, section 3RFC 2119, section 5RFC 8174, section 1

RFC 15

RFC 3935RFC 3629RFC 7322RFC 2418RFC 2026RFC 2028RFC 8126RFC 8179RFC 8729RFC 9280RFC 7841RFC 6410RFC 7282RFC 2119RFC 8174

Specification 9

TR 21.801 v19.1.0TR 21.905 v19.2.0TS 21.101 v19.0.0TR 21.919 v19.0.0TS 21.201 v19.0.0TS 21.202 v19.1.0TS 21.205 v19.0.0TS 21.111 v19.0.0TR 21.802 v1.2.1

Page on the standards body's own site 7

The 3GPP Working ProceduresThe 3GPP websiteThe 3GPP portalThe change-request databaseChange requests, step by stepThe IETF websiteTR 21.900 as a readable web page

Where to go from here

The raised numbers on this page

  1. 47,474 of them are liaison statements sent to another group — rows in table `tdoc` where tdoctype='LS out', read from /var/www/whatthespec.net/data/database/api/api.sqlite on 2026-08-04
  2. 1,629,643 meeting documents recorded — rows in table `tdoc` of the pygppe meeting-document database, read from /var/www/whatthespec.net/data/database/api/api.sqlite on 2026-08-04
  3. The IETF website — the address as written in RFC 2418 section 1, where it is written http://www.ietf.org