Summary

  • NANOG's public 2025 materials expose most stages of a strong amendment record: a dated proposal, exact additions and deletions, a ballot timetable, distinct voting denominators, a third-party voting-service certificate, a membership-adoption date and current text incorporating the changes.
  • The 97.6% result has a precise scope. It is 165 yes votes out of 169 non-abstaining amendment votes, alongside 184 total ballots, 15 abstentions, 725 eligible electors and certified turnout of 25.4%.
  • The sources reviewed do not establish a separately stated effective date or a visible, continuous run of adjacent consolidated versions after October 2017. That public-page link gap says nothing by itself about NANOG's internal or authoritative records and does not call the votes' validity into question.
  • A proportionate response would be a compact, maintained amendment ledger joining proposal authority, redline, aggregate result, adoption, separately stated effectiveness when available, and the immediately prior and adopted texts—without exposing individual ballots, member accounts or protected internal material.

The handoff after a well-documented vote

On NANOG's current bylaws page, two dates perform two different jobs. The page identifies 23 September 2025 as the proposal date for the latest amendments and 5 November 2025 as the date on which the membership adopted them. Farther down, a visible list of “Previous versions” offers older copies, but that list ends at October 2017. A reader can therefore see that the current document has changed since the last version linked in that sequence. What the page does not visibly offer is an adjacent run of consolidated copies carrying the reader from that 2017 endpoint through the amendments reflected in the current text.

This is a small observation with strict limits. It concerns links visible on one public page. It does not show that later versions are absent from every NANOG location, that authoritative copies were not retained, that internal records are incomplete or that any amendment failed to take legal effect. NANOG may hold complete records outside the ten public sources examined here. A nonprofit's reader-facing bylaw page need not duplicate its formal corporate record. The issue is narrower: whether a member or outside reader can reconstruct the public change path without rediscovering scattered pieces each time.

That distinction prevents the story from becoming a cheap transparency verdict. The 2025 public record is detailed. The proposal and election summary expose exact additions and deletions. The page supplies an election window and aggregate participation figures. A dated Simply Voting certificate gives the exact yes, no and abstention counts. The current bylaws identify proposal and adoption dates and visibly incorporate the subjects of the change. Most of the evidence chain is already in public view; its final version-history handoff is the part that remains difficult to follow from the current page.

The institutional scope matters as much as the documentary one. NANOG is a voluntary network-operator forum and nonprofit organisation. It is not a government, a regulator, a regional internet registry, a resource allocator or the operator of the independent networks represented around its community. Its membership can make decisions about NANOG's own bylaws under those bylaws. That corporate decision does not need to express the preferences of every attendee, sponsor, operator or North American internet user.

Public reporting should credit the act at its true institutional scale, neither diminishing the members who voted nor enlarging their constituency beyond the organisation.

Seen that way, the question is not whether the 2025 ballot was “transparent enough” in the abstract. It is whether each public record remains attached to the question it can answer. Who could place the proposal on the ballot? What wording did members consider? Which figures describe eligibility, participation, abstention and choice? What did the voting service certify? When did the membership adopt the amendment? Was effectiveness stated separately? Which consolidated text preceded the change, and which one followed it? The historical trail is strongest when those answers are linked but not collapsed.

One amendment process, several evidentiary jobs

Calling all of these materials “the amendment record” hides their division of labour. A proposal fixes the wording offered and may give the proposer's rationale. A redline shows additions and deletions. Notice and discussion information describe the opportunity for eligible members to consider the measure. A result page records the ballot outcome at whatever level of detail it publishes. A voting-service certificate records the service's aggregate processing and tabulation claims. A corporate adoption date identifies the membership act. A separately stated effective date, if one exists, answers when the new rule began to govern.

A consolidated text shows the bylaw document as a whole after the change. A preserved prior version makes the transition reproducible later.

Each job has a characteristic failure of substitution. A proposal cannot prove adoption. An approval percentage cannot supply an unstated denominator. A voting-service certificate cannot establish the authoritative wording of the bylaws. A current consolidated page cannot show, by itself, which exact version members considered. Adoption may coincide with effectiveness, but the two should not be treated as synonyms unless the record says that they coincide. The sources here establish adoption on 5 November 2025; they do not establish a separately stated effective date.

The faithful entry is therefore “not separately stated in the reviewed public record,” not a copied adoption date.

The current bylaws also distribute authority rather than presenting NANOG as one actor doing everything. Article 11 permits amendment by majority vote of the members. It allows the Board to place an amendment on the ballot and provides a route for a qualifying member petition to do so. That rule identifies possible routes to ballot placement; it does not establish which actor originated every historical proposal.

Article 10 describes annual and interim election routes, at least 30 days' notice for an interim election, an Election Committee, electronic voting, and a member meeting and/or open plenary for presenting and discussing amendments. Those are features of the current text. They are not proof, without ballot-specific evidence, that each listed step occurred in each earlier year.

This role separation is not evidence of conflict. Distributed responsibility can be appropriate. The Board or qualifying petitioners may place a measure before members. The Election Committee has an administrative role under the current rules. Eligible members vote. A service provider may process and tabulate the ballots. The membership adopts the change through the prescribed vote. A text custodian maintains the consolidated bylaws and their public presentation. A reconstructable history assigns each stage to the evidence and actor appropriate to it rather than attributing every action to an undifferentiated institution.

The result is a chain of handoffs, not a demand that one page perform every function. Election pages can remain election pages. The current bylaw page can remain the clearest place to read the governing text. Certificates can remain issued documents. The archival design problem is to give the stages a stable identifier and a path between them. That standard makes the records from 2017, 2018, 2019, 2020 and 2025 comparable without pretending that they all expose the same fields.

2017: exact wording travels farther than an unattached denominator

The archived 2017 proposal is strong where many amendment records are weakest. It publishes exact committee-related changes and explains the rationales offered for them. The subjects include committee structure, staggering terms for standing committees and the organisation of the Program Committee. It also links the then-existing October 2016 bylaws. A reader can begin with a complete prior text, inspect the proposed changes and understand NANOG's explanation. That is unusually useful proposal evidence. It does not, by itself, show that members approved the changes, that the wording was adopted or that it became effective.

The corresponding result page adds a different kind of evidence. At the level of the election page as a whole, it reports 687 eligible voters and 197 total votes cast. In a separate amendment section, it reports 92% approval and publishes adopted section text. The page structure is decisive: the 687 and 197 values are page-level election aggregates. The reviewed material does not establish 197 as the amendment-specific ballot count, so it cannot serve as the denominator for the 92% figure. Nor can that percentage be reverse-engineered into invented yes and no counts.

The amendment section supplies neither an amendment-specific abstention count nor a separately stated effective date.

Restraint here is more informative than forced completion. The public record permits a precise statement—reported amendment approval of 92%, with amendment-specific numerator and denominator not established in the source set. It does not permit a reconstructed count. Marking the missing scope protects the reported result from being blended with figures that describe a broader election page. It also leaves room for another authoritative public record to supply the missing detail later without requiring a correction to fabricated arithmetic.

The 2017 materials therefore illustrate two forms of portability. Exact wording and a prior consolidated text travel well because a later reader can compare them directly. A percentage whose measure-specific denominator is not attached travels less well because its scope can be lost when quoted apart from the original page. That is an information-design difference, not a conclusion about the vote. A valid member decision and an imperfectly portable public summary can coexist.

This is also the first reason an amendment ledger should tolerate genuinely blank fields. Standardisation is valuable only if it preserves uncertainty. A chronological table that forces every year to display a yes count, a no count and an abstention count would create false symmetry. A better record would state “not separately published in the reviewed result” for those 2017 fields, while preserving the 92% exactly as NANOG reported it and linking the adopted section text.

2018: one package exposes the granularity of choice

The 2018 proposal makes a different trade-off visible. The Board said the revisions were intended to improve organisation, readability and consistency with reality, and the page invited comments. Thirty-four sections were presented in a single yes-or-no package. NANOG characterised 19 of those sections as legally significant or containing new or clarifying language after legal review; it characterised 15 as grammatical-only. The page linked the existing bylaws, a draft and the complete proposed text.

That amount of proposal material supports close reading. The package structure, however, means the ballot collected one answer across many edits. Bundling can simplify administration, shorten a ballot and keep an interdependent revision together. It also reduces section-level choice: a voter who favours the grammatical work but questions one clarification, or accepts most structural changes but not one clause, cannot express those preferences separately in the package vote. Describing that choice architecture is not an accusation. The public record examined here does not establish unlawful bundling, deception, manipulation or invalidity.

The useful archival fact is the ballot's granularity. A durable entry should record that 34 sections travelled together, preserve NANOG's 19-and-15 classification as the organisation's own description, and link the exact package. It should not reinterpret that classification as an independent legal judgment or infer why any member voted yes, no, abstained or stayed away. The structure of the question is public; individual motivations are not.

The 2018 results again resist an attractive but unsupported calculation. The election page reports 682 eligible voters and 192 total votes cast at page level. Its amendment section separately reports 93% approval. The source set does not establish that 192 is the amendment-ballot denominator. It therefore supports no derived amendment-specific yes, no or abstention count. Results for Board candidates elsewhere on that page belong to another contest and cannot be merged into the bylaw measure.

Here, as in 2017, a rich proposal coexists with a headline result whose measure-specific denominator is not exposed in the reviewed passage. Yet 2018 adds a version-history reason for preserving complete documents. A list of 34 edits tells a reader what was put to the vote. It is less efficient for learning the complete state of the bylaws after those edits have been applied. The immediately prior consolidated text and the adopted consolidated text answer that second task. One does not replace the other: the redline preserves decision granularity, while the adjacent versions preserve documentary state.

The distinction becomes more important as time passes. At voting time, entities may remember the debate and know which draft is current. Years later, a researcher sees links and labels without that context. A stable package identifier, ballot result and version pair allow the meaning of “the 2018 amendments” to survive beyond the page design in which those materials first appeared.

2019: a simple amendment can still leave a thin chain

The 2019 annual report compresses the relevant episode to a concise institutional update. It says NANOG held elections for Board seats and proposed bylaw amendments. It then says that one change was posted to the bylaws: the official corporate name changed from NewNOG, Inc. to NANOG, Inc. That is the full safe reach of the passage for this purpose.

The report does not expressly state, in that passage, that members approved the name amendment. It gives no amendment tally, approval percentage or abstention count. It discusses broader election-process activity elsewhere, but those general process details cannot all be assigned specifically to the bylaw ballot. The source supports an election-and-proposal context and a posted name change. It does not support an invented approval formula or measure-specific procedural history.

The apparent simplicity of the substance makes this episode analytically valuable. A corporate name change is easier to describe than a reorganisation of committees or a 34-section revision. But a future reader can still ask distinct questions: which clause changed, which route placed it on the ballot, what aggregate amendment result was published, when the membership adopted it, whether effectiveness was stated separately, which full text preceded it and which full text incorporated it. Simple content does not eliminate the need to identify documentary state.

An annual report need not become a transaction register. Its purpose is broader, and its compact language may be entirely appropriate to that purpose. The preservation opportunity is a stable link outward. A one-paragraph annual narrative can point to a small amendment entry containing the transactional record; the amendment entry can point back to the annual account. Neither page must carry every detail.

This is where proportionality becomes practical rather than rhetorical. If the recording method is too elaborate for a name change, it will be applied selectively or retrospectively. A compact set of standard fields collected while the amendment closes can accommodate a one-line change and a complex package without pretending that they are equally consequential. The discipline lies in consistent identity, not equal narrative weight.

2020: a measure-specific result becomes easier to carry

The special bylaw election of 2020 presents a cleaner combination of proposal and result. The ballot ran from 20 to 22 July. Its page published a redline covering the start and end of Board terms, the timing of officer appointments and immediate vacancy replacement. For that bylaw ballot, it reported 583 eligible voters, 125 ballots and 96% approval.

Those figures belong to the special bylaw measure. The same page contains numbers for a separate Board election, but those results answer another question and must remain separate. Keeping the contests distinct is not merely tidy presentation. Without measure identity, a valid aggregate can migrate to the wrong decision and acquire a meaning it never had.

Compared with the earlier archived pages, the 2020 record makes it easier to carry the ballot window, scope, eligible electorate, ballot count and approval percentage together. It demonstrates that result reporting can become more precise even if version preservation remains a separate concern. The sources reviewed do not establish a complete public pair of superseded and adopted consolidated texts for that episode. That limitation concerns the evidence set and visible path, not NANOG's internal retention.

The episode also argues against building one giant amendment page. Different public records can improve independently. An election result can become more measure-specific; a proposal can offer a clearer redline; a consolidated history can gain better labels. The durable design task is to connect those improvements. Stable identifiers and adjacent-version links prevent a strong ballot record from becoming detached when a page hierarchy or current text later changes.

Across the first four episodes, the pattern is not a simple rise from bad records to good ones. Each year exposes a different strength. The 2017 proposal gives exact changes, rationales and a prior-text link. The 2018 package preserves extensive proposal material and reveals decision granularity. The 2019 report captures a concise organisational outcome. The 2020 page attaches measure-specific aggregates to a redline and election window. These pieces supply the vocabulary for understanding why 2025 is unusually complete—and why completeness at the ballot stage still leaves a version question.

2025: four denominators, one bounded decision

The 2025 election summary and results page comes closest to a joined public journey. It explains and marks the precise additions and deletions in three areas: authority over the election method, candidate presentation in an open plenary, and wording about communication platforms. NANOG described the objective as alignment with recent practice. That rationale belongs to the organisation; the public sources do not independently establish that the revisions were necessary, superior or legally required.

The current consolidated bylaws visibly reflect those subjects: Board-set election method, open-plenary presentation and updated communication-platform wording. The current page also identifies 23 September 2025 as the proposal date and 5 November 2025 as the membership-adoption date. The proposal and current text can therefore be read together. What must not be supplied by inference is a separately stated effective date; none is established by the reviewed sources.

The aggregate vote record is strongest when read as four nested scopes rather than as one headline. NANOG reported 725 eligible electors. A total of 184 ballots were cast. Of those ballots, 15 abstained on the amendment question, leaving 169 non-abstaining amendment votes. Those 169 consisted of 165 yes votes and four no votes. The published 97.6% approval is 165 divided by 169. It is not 165 divided by 184, and it is not 165 divided by 725.

The Simply Voting certificate dated 5 November 2025 records the same aggregate structure. It reports 184 ballots from 725 electors and gives turnout as 25.4%. For the proposed bylaw amendments, it records 165 yes, four no and 15 abstentions. It presents yes as 97.6% and no as 2.4% of the 169 non-abstaining amendment votes, while the 15 abstentions are 8.2% of the 184 ballots. The arithmetic closes without changing scope: 165 plus four equals 169, and 169 plus 15 equals 184.

The labels guard against two opposite exaggerations. Calling 97.6% “the membership's view” without its non-abstaining denominator would make the result sound broader than the count. Calling 25.4% turnout evidence that the decision lacked force would make participation perform a legal and institutional judgment that the figure cannot supply. The safe account is direct: a decisive yes share among the 169 members who cast a non-abstaining vote on the amendment, within 184 ballots from an electorate of 725.

The figures reveal nothing about why 541 eligible electors did not cast a ballot. They cannot be assigned to opposition, support, indifference, lack of notice, satisfaction with the proposal or any other explanation. The 15 abstentions are equally bounded. They record ballots that selected neither yes nor no on the amendment question; they do not disclose a reason. Aggregate categories become misleading when a narrative is invented for the people inside them.

NANOG's member-voting guidance contributes an operational but limited piece. It directs members to access ballots through their profiles, review voting-access information and confirm that membership is in good standing. That is useful public guidance about the access route. It is not an audited electoral roll, does not establish the final eligible-elector register and cannot settle a hypothetical eligibility dispute. Instructions about how a voter proceeds and a certificate reporting aggregate turnout answer different questions.

The certificate itself also needs a precise label. Simply Voting says that its independently managed service securely processed and accurately tabulated the results. The document is a third-party voting-service certificate recording the vendor's aggregate tabulation. It is not an independent audit of the election. It does not establish legal effect, the completeness of NANOG's archives, the authoritative consolidated wording, the scope of any review methodology or the preferences of non-voters. Individual ballots remain private.

Those boundaries do not weaken the certificate's value. They make it usable. The dated document fixes what the service reported: electorate, ballots, turnout, amendment yes and no counts, and abstentions. A reader can reproduce the 97.6% denominator instead of guessing. That is materially stronger than an unattached approval percentage. It simply remains different from proof of adoption and different again from the consolidated text.

The 2025 chain can therefore be credited in full. The public materials expose the proposal's subjects and wording, an election timetable, the eligible-elector denominator, total ballots, abstentions, non-abstaining amendment votes, exact yes and no counts, a third-party service certificate, a membership-adoption date and current text incorporating the changes. The visible handoff that remains hard to make from the current bylaw page is between adjacent consolidated versions after the previous-version list's October 2017 endpoint.

Participation evidence has an institutional perimeter

The coexistence of 97.6% approval and 25.4% turnout invites a contest over which number tells the “real” story. Neither can carry that burden. Approval answers how the 169 non-abstaining amendment votes divided. Turnout answers how 184 ballots compare with 725 eligible electors. The abstention figure explains why the ballot denominator and the yes-or-no denominator differ. Reporting all four scopes is not equivocation; it is the only way to avoid turning one numerator into a whole constituency.

NANOG members who participated made an internal organisational decision under NANOG's rules. That is the relevant authority. The decision does not become more valid by being described as an industry mandate, and it does not become less valid because it is not one. A voluntary forum's membership and a continent's network-operating community are not interchangeable groups. The result should be credited without annexing the views of people who did not vote or were never part of the electorate.

This bounded reading also changes the tone of longitudinal comparison. The 2020 result attaches 583 eligible voters, 125 ballots and 96% approval to its special bylaw ballot. The 2025 sources provide the exact arithmetic behind 97.6%. The 2017 and 2018 pages, by contrast, publish page-level eligibility and vote aggregates while reporting amendment approval separately. A responsible table should not force those years into identical columns by borrowing the page totals. “Not separately stated” is evidence discipline, not a mark against the decision.

The same principle applies beyond numbers. A Board rationale is evidence of what the Board said, not proof of necessity. A member-voting instruction is evidence of a stated access procedure, not proof of the final roll. A certificate is evidence of the vendor's aggregate tabulation, not the bylaw text. A current page is evidence of the visible current wording, not by itself a complete history of every prior consolidated copy. Accountability begins by keeping each proposition inside its evidentiary perimeter.

That perimeter is what makes the proposed public record modest. It is not designed to determine whether a vote was socially representative, to review individual eligibility or to revisit the membership's choice. It is designed to let a later reader reconstruct the public act at the level at which NANOG documented it: proposal, class-level participation, aggregate outcome, adoption and textual transition.

Adoption, effectiveness and the state of the document

Ballot closure is not the end of the documentary journey. “Passed,” “adopted,” “effective,” “incorporated” and “current” can point to different moments. A result page can announce the count. The membership can adopt the amendment. The amendment can specify immediate or later operation. A custodian can update the consolidated document. Later amendments can turn that adopted copy into a historical one. If all of those states are reduced to a single date, future readers cannot tell whether the record established coincidence or an editor assumed it.

For the latest amendment, the reviewed sources support proposal on 23 September 2025 and membership adoption on 5 November 2025. They also support that the current text reflects the changes. They do not establish a separately stated effective date. Adoption therefore must not be republished as effectiveness. A durable record can preserve the distinction with a simple value: “not separately stated.” Should another authoritative source later supply an effective date, the field can be updated with a dated correction and link.

The version boundary calls for the same restraint. A visible list ending in October 2017 does not prove that NANOG has no consolidated versions from later years. It shows only that the current page's displayed previous-version path does not lead the reader through them. If a stable public history already exists elsewhere, the best intervention may be a prominent link rather than a new archive. If the versions are retained only internally, then the issue is one of public discoverability and access policy, not proof of loss.

Why preserve a complete prior copy when the redline is already available? The two support different reading tasks. A redline isolates edits and is ideal for understanding the proposition placed before voters. A consolidated prior copy preserves definitions, numbering, cross-references and untouched clauses in context. A consolidated adopted copy shows the whole governing document after the changes were applied. Together, the three allow a reader to inspect the proposition and verify the state transition without manually rebuilding the bylaws.

The smallest durable version path is therefore three-way: immediately prior, adopted and current. At the moment of adoption, the adopted and current links may identify the same document. Later, another amendment can make them diverge. Preserving the adopted copy when the change closes prevents the earlier endpoint from vanishing as “current” moves forward. Descriptive labels and dates should do most of the work; an optional file digest can help detect replacement, but it should not turn a reader-facing history into a technical obstacle.

Version continuity changes the value of the ballot evidence. Without it, a future reader can know that members approved a proposal but still expend considerable effort learning which complete text followed. With it, the redline, result and governing state reinforce one another. The gain is not more authority. It is less interpretive friction.

It also makes later correction more intelligible. If a typographical repair changes a consolidated file after adoption, a stable adopted version lets the custodian explain whether the correction altered substance, presentation or only the index. Without a fixed endpoint, a reader may see two files and have no way to tell whether the difference records another member decision or routine document maintenance. Version identity therefore protects both sides of the boundary: it prevents an administrative edit from masquerading as an amendment, and it prevents a real amendment from disappearing into an undated replacement.

The strongest case for leaving the public architecture alone

A serious recommendation must survive the best objection. Corporate rules derive their force from valid organisational acts and authoritative retained text, not from a public website imitating a source-code repository. Notices, ballots, minutes, legal records and controlled copies may establish a complete internal chain even when a reader-facing link list is selective. The public sources reviewed here do not show a failure in NANOG's internal retention, and they do not establish that another public ledger is legally required.

Disclosure also has a perimeter. Individual ballots should remain private. Named member-account histories, person-level eligibility material, security controls, privileged legal advice and internal deliberations are unnecessary to reconstruct the aggregate public act. Publishing them would create risks that a version index is not meant to solve. Even indiscriminate publication of working drafts can make it harder to distinguish the wording considered, the wording adopted and the text that governs.

Administrative burden is not a trivial defence. Every permanent public register needs ownership, correction practice and link maintenance. A well-designed page can become misleading if nobody updates it, if a file is replaced silently or if labels survive after their targets move. A small nonprofit may reasonably prefer a concise authoritative current page to a sprawling historical system that it cannot maintain. More disclosure is not automatically better recordkeeping.

There is also a risk of status confusion. A convenience index should not compete with the governing text, nor should a certificate be presented as the corporate adoption record. If readers cannot tell which component is authoritative for which purpose, the new layer can worsen the exact problem it was meant to solve. The countercase therefore sets a demanding design test: any addition must be smaller, clearer and easier to maintain than the reconstruction burden it removes.

That test rules out a document dump. It rules out publishing private voter data or protected institutional material. It rules out treating every draft as a version of the bylaws. It also rules out a historical portal whose implied completeness exceeds its maintenance capacity. The case that remains is practical rather than legal: index public, class-level material already produced during the amendment process, preserve adjacent adopted states and make uncertainty explicit.

The smallest useful amendment ledger

A proportionate ledger can fit into a chronological index with one stable entry per amendment package. It does not replace the current bylaws, election pages or issued certificates. It joins them. Ten compact fields are sufficient.

  1. Proposal identity. Give the measure a stable identifier, date and version label. The identifier prevents confusion when several proposals appear in one year or a correction is later attached.

  2. Ballot-placement route. State whether the applicable public record identifies Board action or a qualifying member petition under the bylaws. Record the class of authority, not private drafting history or individual deliberations.

  3. Notice and discussion. List only the public dates and venues established for that proposal. The current bylaws describe notice and presentation mechanisms, but a historical entry should not project today's procedure backward without ballot-specific evidence.

  4. Exact proposal and redline. Link the wording members considered. If provisions travelled as one package, identify the bundle. In 2018, that would include NANOG's own description of 19 significant or clarifying sections and 15 grammatical-only sections, without converting the description into an independent assessment.

  5. Eligibility class and aggregate electorate. Describe the relevant membership class and publish the aggregate eligible-elector figure where available. Keep names, account histories and individual eligibility determinations outside the public entry.

  6. Ballot window and result. Record total ballots, yes, no and abstentions with a scope and denominator for every percentage. For 2017 and 2018, amendment-specific counts not established by the reviewed pages should remain “not separately stated,” rather than being derived from election-page aggregates. For 2025, the record should retain 165 yes out of 169 non-abstaining amendment votes alongside 184 ballots, 15 abstentions and 725 eligible electors.

  7. Published certificate or correction. Link the issued aggregate certificate and any later correction with an accurate description of its source and scope. The 2025 Simply Voting document is a third-party voting-service certificate, not a separate audit of the entire election.

  8. Adoption and effectiveness. Give adoption and effective dates separate fields. Where the public record supplies adoption but no distinct effective date, say so. Never populate two boxes with one date merely to make the table look complete.

  9. Adjacent consolidated versions. Preserve the adopted-text version and link the immediately prior and current copies. Use stable addresses and human-readable labels. A digest can be supplementary, but the version must remain intelligible without specialist tools.

  10. Privacy and authority boundary. State that the ledger contains public governing documents and aggregate election information. Exclude individual ballots, member-account data, security details, privileged advice and non-public deliberations. Identify the current bylaw page as the source for current text and the ledger as an index.

Many of these fields already exist across NANOG's public materials. The proposed work is less about producing new disclosure than about preventing established pieces from drifting apart. The ledger could display only a row and links for a simple amendment, while a large package could use a detail page. Consistency would come from stable identity and scope, not from forcing equal quantities of prose onto unequal changes.

The format must also preserve “unknown,” “not applicable” and “not separately stated” as distinct values. Unknown means the reviewed entry cannot establish the answer. Not applicable means the field does not belong to that amendment. Not separately stated means a related date or count exists, but the specific value was not published as a distinct item in the record. Those labels keep gaps legible without turning them into allegations.

Preservation requires correction and ownership

A ledger without a correction practice simply gives errors a permanent address. A minimal correction method would add a dated note, identify the field changed, retain enough of the earlier entry to understand the alteration and specify whether the governing text changed or only its description. An administrative correction to a label or link must not look like a new bylaw amendment.

Ownership should be assigned by function rather than by a named individual. The election function may maintain proposal and result references; the bylaw custodian may control consolidated versions; a designated organisational function may own the index and link checks. The roles need not be centralised. They need a closing handoff so that publication of the result is followed by preservation of the adopted version and update of the chronological entry.

The least expensive moment to capture the record is while the amendment is active. Proposal identifiers, redlines, dates and ballot authority are already available before voting. Aggregate figures and the certificate arrive with the result. Adoption and the consolidated copy can be recorded when the current text changes. Treating those links as closing materials avoids years-later reconstruction and reduces the temptation to borrow a page-level denominator for a measure whose count was never separately stated.

Maintenance should remain visible but modest. Periodic link checks, immutable version addresses where feasible and dated corrections are more useful than an elaborate system that cannot be sustained. If the index points to an external or legacy archive, the entry can preserve a stable snapshot or replacement link without pretending that every underlying page will remain unchanged forever.

There is a further incentive effect. Election administrators are naturally focused on accurate and timely results. Text custodians are focused on the current governing document. Communications staff are focused on reader clarity. None of those incentives automatically rewards preserving the connection between a ballot and a version years later. A closing checklist makes continuity a small operational duty at the moment when the evidence is easiest to assemble, rather than an unfunded historical project.

A reconstructability test, not a legitimacy verdict

The claim here can be falsified. If NANOG already publishes a stable, dated path that joins every proposal and redline to ballot-placement authority, measure-specific aggregate results, the issued certificate, adoption, separately stated effectiveness and the immediately prior and adopted consolidated versions, then the case for a new ledger largely disappears. If such a path exists but is difficult to find, discoverability may be the only problem. If the visible previous-version sequence is expanded, the present concern can become historical rather than current.

Another approval percentage would not answer the version question. Another current consolidated copy, on its own, would not connect the ballot to the state transition. The test is the join. Starting from one amendment entry, can a reader identify the wording offered, understand the route to the ballot, read each aggregate with its proper denominator, distinguish the voting service's certificate from the membership's act, find the adoption date, see whether effectiveness was stated separately, and retrieve both the immediately prior and adopted complete texts?

Can that be done without seeing an individual ballot or private account record?

The historical comparison shows that NANOG has already published nearly all of those components at one time or another. In 2017, exact changes, rationales, a prior bylaw link, a reported percentage and adopted section text were visible. In 2018, the full bundled proposal and NANOG's classification of its sections were visible alongside a reported percentage. The 2019 annual report recorded that the name change was posted to the bylaws. The 2020 special election attached measure-specific aggregates to a redline and dates.

The 2025 materials added exact denominator closure and a third-party voting-service certificate; the current page then identified proposal and adoption dates and displayed the revised language.

That sequence supports no allegation of illegality, inaccurate counting, concealment, manipulation, deficient retention or invalid adoption. It supports a narrower design inference. Public ballot evidence has become richer, but its long-term value depends on whether the document state before and after each amendment remains easy to retrieve. A concise index can preserve that value without widening disclosure into private voting or internal legal material.

The appropriate standard is reconstructability with boundaries. Label the authority and the measure. Preserve the exact proposal and adjacent consolidated texts. Keep every percentage attached to its numerator, denominator and scope. Distinguish a vendor's aggregate certificate from adoption and distinguish adoption from effectiveness. Record uncertainty rather than filling it. Assign maintenance and correction. Leave individual choice and protected institutional records private.

NANOG's 2025 materials show that this standard is already close. The proposal, aggregate decision and current wording are visible in substantial detail. The durable gain would come at the handoff: making sure that when today's current bylaws become tomorrow's prior version, the public trail through the change remains intact.

Sources