Daily Dose of 6G

The study is over. Now someone has to write “shall.”

TS 22.270 is where 6G stops being a description and becomes an obligation. The document exists today, as a draft, and the group that has to finish it meets tomorrow in Prague with three meetings left before the Stage-1 freeze.

Transcribed from 3GPP’s published Release 21 timeline (© 3GPP 2026). Highlighted row: the Stage-1 freeze that TS 22.270 has to make.

What a TS costs that a TR does not

Editions 03 and 04 covered TR 22.870, the SA1 study: six use-case families, AI agents written in as functional requirements, roughly two years of contributions from ninety-plus companies. A Technical Report is allowed to be interesting. It can describe a use case, note that the network “may” need to do something, and leave the question of exactly what open. Nobody is ever held to a TR.

A Technical Specification is a different instrument. Stage 1 in 3GPP’s three-stage model is the service description — what the system does, stated from the user’s side, in language that Stage 2 architects and Stage 3 protocol designers are then obliged to satisfy. The verb changes from may to shall, and every shall creates downstream work for someone in SA2, CT1 or RAN2 who will have to build it and prove it.

That is the whole difficulty of the next seven months. The study said 6G should support AI agents. The specification has to say precisely which capability the 6G system shall provide, to whom, under what conditions — in a sentence an implementer cannot argue their way out of, and a test house can eventually check.

The document that exists today

It is worth being exact about what is and is not on the record, because the gap between the two is the story. The 3GPP portal specification entry, checked this morning, reads:

Two things there are worth pausing on. First, the pen is shared across an operator pair that spans both sides of the Pacific — and Jean Trakinat (T-Mobile USA Inc.) is also co-rapporteur of TR 22.870 alongside Xiaonan Shi (China Mobile). The same hand carries the study into the specification, which is the mechanism by which requirements survive the crossing intact rather than being relitigated from scratch.

Second, a two-digit minor version in June, against a March 2027 freeze. That is not alarming — a v0.x at this stage is normally a skeleton with agreed clause headings and a handful of landed requirements — but it does mean the substance of Rel-21 Stage 1 has still to be written, in the open, in the next three meetings.

The anchor, and what an anchor means

3GPP’s own Release 21 page states the relationship plainly: TSG SA approved TR 22.870 in Rel-20 and it “is set to be the Stage 1 anchor for Rel-21 normative work.” Anchor is the right word and a limited one. It fixes the starting point; it does not guarantee the destination. Requirements that could not attract consensus as a should in a study very rarely survive promotion to a shall in a specification, and the study’s more speculative material — the parts that read as vision rather than requirement — is exactly what gets trimmed in this phase.

Three meetings, then the door closes

SA1’s meeting calendar makes the arithmetic concrete. Between today and the March 2027 Stage-1 freeze there are three ordinary working-group meetings in which text can be agreed:

Then TSG#115 in Q1 2027 approves the Rel-21 package and freezes Stage 1. SA1#117 finishing four weeks before that plenary is not a coincidence; it is the last chance to land text with enough margin for the plenary paperwork.

How to read a v0.x specification

If you go and open the draft, read it structurally rather than literally. The clause headings tell you which requirement areas SA1 has agreed exist — that alone is signal, and it is usually settled well before the text under them is. Empty or one-line clauses mark territory that is claimed but unresolved. Editor’s notes are where the actual disagreements live, and they are the most informative sentences in any 3GPP draft. And any requirement carrying a bracketed value is a number nobody has agreed to yet.

Do not quote requirement numbers from a v0.x outside the group. They renumber between versions, and a requirement cited from a draft that later moves or disappears is how incorrect 6G claims get into circulation.