Summary

  • The IGF Policy Network on Open Digital Infrastructure held its second planning meeting on 2 September and discussed a shortened path to its first stock-taking draft after the consultant's start moved later than expected.
  • Participants proposed a 16 September end-of-day Anywhere on Earth cutoff for initial inputs, with design meetings in the following two weeks and an open network meeting in the week of 28 September.
  • The meeting summary assigns the consultant two distinct functions: synthesize community contributions and conduct research while adding suggestions and insights. Final content decisions remain PNODI's responsibility.
  • Daniel Kade proposes two source lanes inside the draft record: one for participant-originated material and one for consultant-originated evidence or proposals, followed by a shared disposition and revision history.
  • This is an editorial proposal, not a PNODI rule. The checked record does not identify the consultant, confirm the start date or show that the proposed milestones have been completed.

One person will hold two pens

The first draft of an institutional report often looks more settled than the process that produced it. Paragraphs arrive in one typeface. Citations sit in one sequence. Similar proposals are combined, awkward disagreements are shortened and missing context is supplied. By the time readers encounter the document, they can no longer see which sentence came from an open consultation and which was added by the person paid to make the consultation readable.

That risk is unusually concrete in the second meeting summary of the IGF Policy Network on Open Digital Infrastructure. PNODI met online on 2 September 2026 at 14:00 UTC to amend its work plan and review input for an initial stock-take. The consultant whom the network expected to begin in early or mid-August had not yet started. Participants expected work to begin the following week, and the action list asked the IGF Secretariat to confirm the hiring date by 3 September.

The schedule therefore tightened. A proposed initial-input deadline fell on 16 September at the end of the day Anywhere on Earth. A first design meeting was planned for the week of 14 September, another for the week of 21 September, and an open network meeting for the week of 28 September. The indicative path then moved toward a second draft in mid-October, public comment around 20 October to 20 November, and later finalization.

These are proposed milestones, not completed acts. The checked public record does not name the consultant, contain a signed engagement or show a draft. The delay is not evidence of misconduct. It is important because it concentrates several editorial functions in a short interval.

The summary says the consultant should work through the mailing-list archive and community contributions, develop an outline and prepare the draft. It also says the consultant is expected to undertake research and contribute suggestions and insights. Finally, it leaves responsibility for final content decisions with PNODI.

Those sentences describe three legitimate forms of agency: contributors supply material, a consultant synthesizes and adds, and the Network decides. A credible draft should not flatten them into one anonymous institutional voice.

A synthesis is already a set of decisions

Open digital infrastructure is not one easily counted object. At PNODI's first meeting on 27 July, participants discussed open-source software, open standards, protocols and other shared resources. They warned that Digital Public Infrastructure, Digital Public Goods, open-source software and open standards are related but not interchangeable. Governance can mean a maintainer succession problem in one project, a standards-process question in another, a public procurement dependency in a third and a funding model somewhere else.

A stock-take must therefore classify. It must decide what enters the map, which cases can be compared, how funding differs from sponsorship, whether a software library and a protocol share a category, and when evidence from one jurisdiction travels badly to another. Even a faithful summary chooses a boundary, a level of detail and a vocabulary.

That is not a defect. Without synthesis, a mailing-list archive remains a pile of messages rather than a useful public output. The problem begins when an editorial decision loses its author and is later read as a collective conclusion.

Suppose three contributors describe underfunded maintenance, while the consultant adds research on procurement assurance. The draft may reasonably connect the two. But participants called for procurement rules would be stronger than the evidence if no participant did. Conversely, if several people propose incompatible definitions of open infrastructure, a single polished definition may hide rather than resolve the disagreement.

The consultant's own research can be valuable precisely because an open call will be uneven. Some regions, languages and technical communities will contribute more than others. Some claims will arrive without evidence. Existing studies may prevent duplication. Professional synthesis should fill factual gaps and expose weak comparisons. It should not borrow the mandate of the submissions it supplements.

Two lanes, one draft

PNODI does not need a cumbersome parallel publication system. It needs a small origin record attached to the working draft.

The community lane should give every material contribution a stable reference, date and channel. It should preserve the contributor's own subject label, an immutable copy or hash of the submitted material, evidence links and any capacity the contributor voluntarily declares. It should mark whether the contribution was public, anonymized or restricted for a stated reason. A mailing-list post proves that somebody contributed; it does not prove that the sender represents a whole country, industry or community.

The consultant lane should record material that did not originate in the consultation. Each item needs a cited source and a simple type: external evidence, comparative analysis or consultant proposal. The label should follow the material even when it is rewritten into the report's house style. Consultant-originated is not a warning label. It is an authorship fact.

Between the lanes sits a merge record. It should show which source items support a paragraph, whether multiple inputs were consolidated, what distinction was retained and whether contrary material exists. If several near-identical messages reflect a coordinated campaign or one cross-posted source, the record should prevent repetition from masquerading as independent confirmation. If genuinely independent cases converge, it should preserve that fact too.

Every item then needs a draft disposition: included, summarized, deferred, outside scope, duplicate, unresolved or awaiting evidence. A short reason is enough. The record should point to the version in which the decision occurred and preserve later changes rather than overwriting them.

The scheme is intentionally modest. It does not ask for private correspondence, contributor rankings or a public argument over every comma. It asks readers to be able to tell whether a proposition entered through participation, professional research or a later institutional decision.

Design meetings need a return path

The meeting summary distinguishes open network meetings from design meetings. The full network provides feedback in the former; detailed development happens in the latter. That division can be sensible. A drafting group small enough to work line by line can resolve citations and structure faster than an open call.

But the design room is also where provenance can disappear. A paragraph may be rewritten to reconcile two views, a new example may be inserted, or a contested term may become the heading for an entire section. The next open network meeting may see only a cleaner draft and no trace of the choice.

Each material design-meeting change should therefore carry a short receipt: the passage affected, the source items considered, the proposed alteration, the reason, unresolved objections and the resulting draft version. When the draft returns to the network, members should be able to review those receipts alongside the text. Approval can then mean approval of identifiable editorial choices, not acquiescence to a document whose construction is invisible.

This also protects the consultant. If a network decision changes or rejects an analytical recommendation, the record shows that the consultant did not silently contradict their evidence. If PNODI retains the recommendation, the final output can state that the Network accepted it rather than pretending the idea came from all participants at the outset.

Open channels need bounded transparency

PNODI's public page says work will proceed through consultations on an open mailing list. The 2 September summary treats that list as the primary input channel. Direct contact with the consultant may be possible, but contributors were encouraged to use the list so members receive material at the same time. Meeting summaries are published, while recordings can be requested.

That is a sound default for common access. It does not remove the need for privacy judgment. A maintainer may want to describe a fragile dependency without exposing an attack surface. A public official may need to distinguish personal analysis from an authorized government position. A small project may reveal a succession risk without publishing the maintainer's private contact details.

The source record can minimize personal data while keeping the process legible. It can show an input reference, origin class, broad contributor capacity when volunteered, evidence status and disposition without reproducing an email address or an undisclosed affiliation. Restricted evidence can carry a custodian, reason class and review date. Anonymous or confidential material should not disappear into the draft as though it were ordinary public consensus; its limits should travel with it.

The same discipline applies to absence. A short input window will not reach every affected community. Silence from a region or stakeholder group is not consent. The stock-take should publish the coverage it achieved, the outreach it attempted and the important gaps that remain.

The Network owns the final verbs

The IGF description of Policy Networks calls them open, multistakeholder and consultative. MAG members facilitate them, and the Secretariat supplies neutral coordination and support for final outputs. The structure can evolve with participating stakeholders. That is a process architecture, not a transfer of decision power to a contractor.

The Meeting II summary gets the central allocation right: final content decisions remain with PNODI. The useful next step is to make those decisions observable. For each disputed or materially altered item, the final record should state whether the Network accepted, revised, deferred or rejected the consultant's treatment. It should identify the meeting or review round in which that happened and link the final passage back to its sources.

Here the wider IGF boundary matters. The forum's own institutional account says it does not produce negotiated outcomes. Its work informs and inspires actors that hold public or private policy-making power. A PNODI report may organize evidence, compare models and recommend next work. It does not become binding because a consultant drafted it or because an open network discussed it.

Lu Heng's Multi-Stakeholder Mirage provides the warning: participation cannot be silently upgraded into authority over those absent. His Minimum Initial Specification offers the practical response: begin with a narrow, verifiable coordination object and leave later decisions with the actors competent to make them.

Applied here, the minimum object is not a universal definition of open digital infrastructure. It is a lineage record that preserves who supplied what, who added what, who changed it and who finally accepted the wording.

A deadline can sharpen the record

The compressed calendar creates a temptation to postpone provenance until after the first draft. That would reverse the economics. Origin is cheapest to record when material enters. Reconstructing it after consolidation requires searching mail archives, comparing versions and asking people to remember why a paragraph changed.

A small set of fields can begin immediately. New contributions receive references as they arrive. The consultant labels external research at first use. A change log accompanies each design meeting. The open network reviews material decisions rather than every keystroke. Corrections append a new state instead of erasing the old one.

This record will not settle every argument about representation. It will reveal where the argument belongs. Readers can distinguish evidence from recommendation, frequency from authority, professional synthesis from participant endorsement and a PNODI choice from an IGF-negotiated outcome that does not exist.

The first stock-take may be useful even if its coverage is incomplete. It may identify incomparable governance models, underfunded maintenance and existing studies that deserve reuse. Its legitimacy will depend less on presenting one frictionless voice than on showing how several voices became a document. Two source lanes are a small price for keeping that transformation honest.

Sources