Summary
- NANOG's 2018 package disclosed changes across 34 bylaw sections, from obsolete-transition and grammatical cleanup to membership discipline, Board vacancies, committee control, staff authority and election rules. The annual ballot separately asked members to choose two directors, but offered only one yes-or-no question on all the amendments.
- A package can protect coherence, reduce voter fatigue and lower administrative cost. It can also be accompanied by a complete redline and still record only acceptance or rejection of the bundle—not clause-level preference, intensity or trade-off.
- The better-documented 2025 package—165 yes, four no and 15 abstentions among 184 ballots from 725 electors—shows what exact denominators can clarify and what they cannot. The practical reform is a package manifest explaining propositions, dependencies, severability and authority shifts, not compulsory atomisation of every edit.
The sentence that defined the ballot
On 18 September 2018, NANOG's Board recorded the decisive feature of that year's bylaw vote. Its electronic-vote minute said the annual ballot would let members select two Board candidates and cast “a Yes/No vote on all the bylaw changes.” In that sentence, a many-part proposal became a binary decision. Members could differentiate between candidates in the director contest. On the amendments, they could move the disclosed package or leave it in place.
The distinction needs precision. This was not an election containing only one question: the ballot also asked members to choose two directors. Nor did members cast 34 amendment votes. NANOG's official proposal listed changes to 34 sections, then placed the entire list beneath one question asking whether the bylaws should be amended as described. Thirty-four disclosed items therefore produced one measured package preference, not 34 measured clause preferences.
That fact does not make the process secretive. NANOG published a detailed inventory and a redline comparing the October 2017 bylaws with the August 2018 proposal. A reader could inspect the proposed language. But disclosure and choice granularity answer different questions. Disclosure asks whether the parts can be seen. Granularity asks whether one part can be accepted while another is rejected. A fuller redline improves the first without adding another control to the second.
This is the central one-bit problem. A yes-or-no package result can establish which of two whole-package outcomes prevailed. It cannot reveal a preference vector inside the package. That informational limit holds whether the bundle was sensible or clumsy, whether the margin was narrow or overwhelming, and whether the resulting text was validly adopted. The appropriate inquiry is therefore not whether the vote “counted.” It is what the vote authorised inside NANOG, what information it preserved, and what information its architecture never collected.
The authority was corporate and bounded
NANOG describes itself in its current bylaws as a Delaware nonprofit corporation and a facilitator of technical discussion, learning and communication, rather than a network operator. A member vote on its bylaws changes NANOG's internal corporate rules. It does not bind a nonmember's network, turn conference attendance into corporate authority or establish a mandate from every operator in North America.
Current Article 11 permits amendments by majority vote of members during an annual or interim election. The Board may place an amendment on the ballot; so may a petition signed by at least 30 members or one per cent of members, whichever is greater. Current Article 10 says annual elections cover Board candidates and proposed amendments, while an interim election may cover amendments or director removal. It assigns an Election Committee to administer electronic voting for at least 48 hours. Those provisions describe the framework that exists now.
They illuminate the resulting system but should not be silently backdated where historical documents differ.
The electorate is also bounded. NANOG's membership policy describes individual membership, dues and a one-year membership term following application, and says members in good standing may vote. Its voting guide instructs a member to check good-standing status in the profile before using Vote Now. These are access conditions, not evidence that every eligible member voted.
Consultation occupies a separate institutional layer. Mailing-list entities, meeting attendees, staff and directors may inform, criticise or develop a proposal. Their contributions can improve the text and expose risks. The recorded authorisation event, however, is the eligible members' ballot. Treating consultation as adoption would overstate informal participation; treating it as worthless would ignore the information it can contribute. The two mechanisms do different work.
A visible precedent, without a usable denominator
NANOG had already used bylaw revision to bring written rules closer to organisational practice. In 2013, it described a rewrite intended to make the bylaws better organised, more readable and more consistent with reality. The archived rewrite page published three dated drafts—25 August, 15 September and 24 September—and invited comments through the members list or directly to the Board. The public page does not report how many comments arrived, what they said or how they were resolved.
The 2013 election summary divided the rewrite into introductory Articles 1–3, organisational Articles 4–11 and legal Articles 12–16. NANOG identified substantive changes involving member terminology, officers, good standing, ad-hoc committees and member meetings. It separately attributed several provisions concerning Board terms, written consent and officer selection to attorney advice. Those descriptions belong to NANOG. They do not license an independent conclusion that every cited provision was legally necessary.
The 2013 amendments were posted on 20 September, voted on from 7 to 9 October and reported approved. The migrated summary publishes no eligible-voter total, total-ballot figure, yes or no count, or abstention count. No approval percentage can responsibly be reconstructed from it. The omission foreshadows a recurring problem: text may be visible while the arithmetic needed to reproduce a result remains absent.
For the focal package, the immediate baseline was the bylaw version adopted in October 2017. The official redline labels that baseline ADOPTED October 2017 and the new text PROPOSED August 2018. The before-and-after comparison makes the proposal unusually auditable at the textual level. It still says nothing about the number of independent choices on the ballot.
How the 2018 package reached the members
The available Board records show a sequence, but not every deliberation within it. On 3 August 2018, the Board held a working discussion about bylaw and policy updates, with legal counsel listed as a guest in the meeting record. On 17 August, it discussed the bylaws again, directed that the document be updated and planned electronic Board review, according to a second meeting record. These documents show counsel's presence during part of the process and continued Board editing. They do not show that counsel drafted, endorsed or approved every clause.
NANOG's proposal page said the organisation had contracted legal counsel to review its corporate documents and address governance gaps. It described 19 of the 34 changed sections as having legal significance or containing new definitions or language, while 15 contained grammatical corrections only. That 19/15 split is NANOG's published classification. It should not be converted into an independent legal judgment, or into a claim that the 19 formed 19 separately approved propositions.
The proposal invited comments and suggestions through the members mailing list or directly to the Board. This establishes a route for participation. The reviewed public sources do not establish the consultation's exact start date, the number or content of responses, the Board's replies or a change ledger connecting comments to final clauses. The absence of a public ledger does not prove silence. It leaves the consultation-to-text path unresolved.
Even the dates need to remain unpolished. The contemporary election schedule says the amendments were published and the ballot approved on 17 September. The formal electronic-motion record says the ballot motion passed on 18 September. Neither source explains the one-day difference. It would be tempting to invoke a timezone, posting delay or clerical slip, but the record does not support any of those explanations. The public timeline says 17 September; the formal motion is dated 18 September.
The 34-section entity, reconstructed
The proposal's ballot text contains 32 numbered change instructions. One instruction created three separately numbered sections—6.6, 6.7 and 6.8—so the official affected-section count is 32 - 1 + 3 = 34. This accounting guards against two errors: reporting only 32 because the instruction list stopped there, or implying that the 34 affected sections appeared as 34 ballot questions.
The inventory below follows NANOG's official list. The brief descriptions show the operating surface of each item without purporting to classify its legal effect independently.
| # | Section | Disclosed change and operating surface |
|---|---|---|
| 1 | 4.1 | Removed Working Groups from organisational and Board-delegation language. |
| 2 | 4.2 | Made grammatical corrections. |
| 3 | 4.3 | Made grammatical corrections. |
| 4 | 4.3.1 | Removed the Term Limits heading. |
| 5 | 4.3.2 | Deleted the expired interim procedure introduced after the 2013 change in director term length and used for the 2014 and 2015 elections. |
| 6 | 4.4 | Added a reference to the definition of good standing. |
| 7 | 4.5 | Made grammatical corrections. |
| 8 | 4.6 | Reworked the definition of a Board vacancy and the procedures governing one. |
| 9 | 4.7 | Removed the first paragraph and cleaned the language governing removal of an elected director. |
| 10 | Article 5 | Used NANOG and aligned officer language. |
| 11 | 5.6 | Made a grammatical correction. |
| 12 | 6.1 | Defined Members in Good Standing. |
| 13 | 6.4 | Added the right of a member, if elected, to serve on the Board. |
| 14 | 6.6 | Added a provision for membership termination. |
| 15 | 6.7 | Added a provision for membership suspension. |
| 16 | 6.8 | Added expulsion and suspension procedures, including notice, an opportunity to be heard, a Board decision and a time limit for challenges in the proposed text. |
| 17 | Article 7 | Required at least one Board member on each standing or ad-hoc committee. |
| 18 | 7.1.1 | Made a grammatical correction. |
| 19 | 7.2 | Added the Election Committee reference. |
| 20 | 8.1.1 | Made a grammatical correction. |
| 21 | 8.2 | Replaced the Communications Committee with NANOG staff, overseen by the Executive Director, as administrator and minimal moderator of the NANOG list. |
| 22 | 8.2 subsections | Corrected subsection numbering and made grammatical changes. |
| 23 | 8.3 | Made grammatical corrections. |
| 24 | 10.1 | Made a grammatical correction. |
| 25 | 10.2 | Named amendments and director removal as interim-election purposes and reduced notice from 60 to 30 days. |
| 26 | 10.3 | Named the Election Committee, adjusted its composition and preserved electronic voting of at least 48 hours. |
| 27 | 10.3.1 | Made grammatical corrections. |
| 28 | 10.3.3 | Added a method for assigning differing seat-term lengths according to election-result order. |
| 29 | Article 12 | Made grammatical corrections. |
| 30 | 13.1 | Made grammatical corrections. |
| 31 | 13.2 | Made a grammatical correction. |
| 32 | Article 14 | Updated nondiscrimination language on counsel's recommendation. |
| 33 | 15.2 | Made grammatical corrections. |
| 34 | Article 16 | Made grammatical corrections. |
The package was heterogeneous even under NANOG's own description. It included expired-transition cleanup, grammar and style, conforming references, institutional mechanics and provisions concerning rights or authority. Deleting section 4.3.2 removed a transition that had governed the 2014 and 2015 elections; it was not a new vote on director term length. Removing Working Groups from delegation language was organisational cleanup. Neither item alone demonstrates a transfer of member rights.
Other items changed consequential operating surfaces. Sections 4.6 and 4.7 addressed the definition and handling of Board vacancies and the removal language for elected directors. Sections 6.6–6.8 established rules for terminating or suspending membership and a process involving notice, an opportunity to be heard, a Board decision and a challenge deadline. Article 7 required a Board member on every standing or ad-hoc committee, while section 7.2 expressly named the Election Committee. These were not interchangeable with a spelling correction.
Section 8.2 changed the named actor administering and minimally moderating the NANOG list: the Communications Committee gave way to staff overseen by the Executive Director. Section 10.2 reduced interim-election notice from 60 days to 30 and linked interim elections to amendments or director removal. Section 10.3 named and adjusted the Election Committee, and 10.3.3 supplied a vote-ranking method for allocating different seat terms. Counsel-recommended nondiscrimination language in Article 14 travelled with these election provisions, membership-discipline rules and grammatical edits.
This mixture is the reason the 19/15 split cannot answer the consent question by itself. Two provisions may both be described as legally significant while affecting unrelated authority surfaces. A seemingly clerical cross-reference may be necessary to make a substantive definition work. Seriousness and dependency are not synonyms. The public materials provide the redline and the list, but no proposition-by-proposition severability analysis. They do not support a finding that every item could safely stand alone. They equally do not support a finding that all 34 had to share one fate.
What the 2018 result does—and does not—measure
Voting opened on 1 October 2018 at 08:00 Pacific Time and closed on 3 October at noon: a 52-hour window. The migrated 2018 election summary reports 692 eligible voters, 192 total votes cast and package approval of 93 per cent. Its visible heading says the Board candidates were elected in October 2019, but the page title, candidates, voting window and amendment result all identify the 2018 election. The heading is a year error within the record, not a reason to relocate the election.
The legacy result page also reports 192 total ballots and 93 per cent approval, but gives 682 eligible voters—ten fewer than the migrated page. Neither source explains the discrepancy. Following the migrated record, 192 / 692 is about 27.75% overall election participation. Using the legacy figure only to expose the conflict, 192 / 682 is about 28.15%. The two figures should neither be averaged nor silently harmonised.
The more important limitation is the missing question denominator. Both pages report 192 total election ballots, but neither expressly says that all 192 answered the amendment question. Neither supplies exact amendment yes, no, abstention or blank counts. Multiplying 192 by 93 per cent would manufacture a vote count and assume a denominator that the sources do not publish. No exact yes count can be recovered. The exact number and share of eligible members who voted yes in 2018 remain unknown.
That is not a technical quibble. Different questions and different events can have different denominators. NANOG's 2020 election record reports 583 eligible voters and 125 votes for a special bylaw election, then 553 eligible voters and 176 votes for the later Board election. Those event-specific figures cannot fill the 2018 gap. They demonstrate why an election-wide ballot total cannot automatically become a question-level response total.
The final October 2018 bylaws are marked adopted on 3 October. They contain the resulting vacancy, membership-discipline, committee, staff and election provisions. The October 2017 baseline, August 2018 redline, Board ballot action, public result and final text therefore form a before/proposal/ballot/result/after record chain. The chain establishes adoption and textual effect under NANOG's reported process. It does not recover the missing amendment counts or transform a package result into clause-level preferences.
The strongest case for bundling
The case against crude bundles is easy to state. The serious case for them deserves to come first. Bylaws are an interacting instrument, not a stack of independent slogans. A changed definition may require conforming references. A newly created authority may require a procedure, decision-maker and deadline. If linked provisions are separated on the ballot, members could approve an operative clause while rejecting the language that makes it workable, producing contradiction or incompleteness.
A complete redline can also reduce comprehension risk. Thirty-four separate questions would not guarantee 34 careful judgments. They could increase fatigue, fragment the document's logic and make dependencies harder to see. Each additional question adds administrative work in ballot construction, testing, instructions, certification and result reporting. A package may therefore lower transaction cost while presenting the amendment as a coherent whole.
There is a further institutional advantage. A single package gives the sponsor and electorate a determinate fallback: if the amendment fails, the pre-existing instrument remains in force rather than a partially amended version assembled by the interaction of many results. That simplicity can matter when some provisions are definitions, some are procedures and others are conforming references. The package becomes a checksum on the drafting exercise—either the integrated proposal takes effect or it does not.
The cost of that clarity is that the fallback operates at the outer boundary of the package, even when a narrower subset might also have been coherent.
The counsel-review context strengthens that defence within its limits. NANOG said counsel reviewed its corporate documents to address governance gaps, and the Board records place counsel at part of the working process. Coordinated review can reduce drafting and consistency risk. What it cannot establish from this public record is counsel's view of every clause, or whether each item was inseparable from every other.
Members also had meaningful counterweights: the proposal and redline were public, comments were invited, the entire package could be rejected, and later amendments remained possible. Those features are strong evidence against characterising the changes as hidden or imposed without a ballot. The ability to reject everything is real control. A later correction route can address an unforeseen problem.
None of those advantages restores the preference information a single question does not collect. Rejecting a whole package is blunt where its components are distinguishable. A later correction consumes time and agenda space. A transparent redline says what will change; it does not let the voter record different answers to different changes. Bundling may be justified, but its practical merits do not repeal its informational limit.
The right test is dependency, not an abstract presumption that packages are either proper or illegitimate. Textual dependency exists when a definition, number or cross-reference must change with another provision. Operational dependency exists when a power depends on a specified actor, procedure or deadline. Policy dependency exists when several clauses implement a common design even though each sentence could remain textually intact on its own. The first two may be demonstrable from the instrument. The third requires an institutional explanation; a common purpose does not make inseparability self-evident.
That test also clarifies what a package sponsor must decide. One task is to identify combinations that would make the document internally inconsistent: a cross-reference without its definition, for example, or a procedure without the office empowered to use it. A second is to identify combinations that remain legally legible but frustrate the intended policy. A third is to identify changes that are merely convenient to process together. All three can support grouping, but they do not carry the same weight. Treating them separately allows members to distinguish necessity from prudence and prudence from administrative economy.
The 2018 list suggests where such an explanation could have helped. The good-standing definition in 6.1 and the cross-reference in 4.4 appear textually related. The termination, suspension and process provisions in 6.6–6.8 form an operating sequence. Their necessary connection to deletion of the expired transition in 4.3.2, however, is not evident from the published materials. Nor do those materials explain why membership discipline and the change in list administration under 8.2 had to rise or fall together. The point is not that no relationship existed. It is that the record does not state one.
There is a middle course between one question and 34. Closely linked discipline provisions might form one group, election-administration changes another, and editorial or conforming changes a third. Grouping can still conceal disagreement within a cluster, and too many clusters can exhaust voters. The relevant standard is proportionality: the ballot's granularity should follow demonstrated dependencies closely enough to preserve coherence without treating the package boundary as inevitable.
Crucially, dependency must be explained before the vote. A 93 per cent package result demonstrates the reported response to the package that was offered; it does not prove the coupling was necessary. The fact that the adopted provisions later operated together likewise does not prove that no smaller coherent combination was available. High approval may validate the choice between the two outcomes presented. It cannot validate alternatives the ballot never presented.
Agenda-setting and adoption are different powers
The Board's role was to develop, group, explain and approve the ballot. That is agenda-setting power: deciding which changes travel together and how the choice is framed. Members then exercised the formal package choice. Board sponsorship is not member adoption; member adoption does not disclose the reasoning behind the Board's grouping.
The result likewise does not measure enthusiasm or opposition for any individual clause. It supplies no evidence about the internal composition of any ballot choice, and this article does not infer one. Formally, many possible clause-level preference patterns collapse into the same package answer. The lost information follows from the ballot architecture, not from speculation about what a particular voter thought.
This separation prevents two opposite mistakes. Declaring the result meaningless would ignore the member act and the operative text that followed. Treating 93 per cent as independent approval of every clause would claim information the vote never collected. The correct reading is narrower: NANOG reported strong approval for the offered package, under its process, without publishing the amendment-question denominator or a clause-level measure.
Participation must remain distinct as well. Inviting comments created an avenue for influence before voting. It did not disclose the number or content of comments, demonstrate that feedback changed the proposal or substitute for the member ballot. Consultation adds context around authorisation. It is not the recorded authorisation itself.
The 2025 comparison: fewer changes, better arithmetic
Seven years later, NANOG used the same basic consent architecture on a smaller package and published a far more reproducible result. The comparison is controlled in one important respect: several disclosed changes still travelled as one certified amendment question. The number of affected provisions and quality of denominator reporting changed; the binary package choice did not.
The precursor came on 23 May 2025, when the Board adopted ranked-choice voting for future Board elections following an Election Committee recommendation, according to the public Board minutes. The later amendment proposal said the bylaw changes would align the written rules with recent practice. It disclosed three groups affecting five section or subsection identifiers.
The first group amended section 10.3.3 to permit ranked-choice methods, including variants such as single transferable vote, allow voters to rank candidates where appropriate and make the Board responsible for setting the tabulation method before nominations began. The second amended 8.1.1 so Board-candidate presentations and bylaw discussion could occur at a members meeting and/or an open plenary, rather than only under the former members-meeting formulation.
The third group retitled 8.2 from Mailing List to Community Communication, expanded its subject from one list to community communication platforms, retained administration by staff under the Executive Director and allowed the Board to appoint an ad-hoc moderation committee. Related changes to 8.2.1 and 8.2.2 applied eligibility and acceptable-use language to those platforms. The five affected identifiers were therefore 10.3.3, 8.1.1, 8.2, 8.2.1 and 8.2.2. The proposal summary prints 8.21 once, but its detailed text and the final bylaws establish 8.2.1 as the correct reference.
These were three disclosed change groups—Board-election tabulation, the candidate forum and community communication—not three separately certified amendment questions. The certified ballot presented one Proposed Bylaws Amendments yes-or-no question. The October announcement archive shows that members were asked to review the proposal, but does not reveal how many did so, what feedback was submitted, a formal comment deadline or whether comments prompted revisions.
The current bylaws state that the latest amendments were proposed on 23 September 2025 and adopted by the membership on 5 November. They contain the ranked-choice, open-plenary and communication-platform provisions described in the proposal. The proposal and final text therefore establish the intended and resulting change. They do not separate member approval by clause.
The date record contains its own tension. The 2025 election summary has a header saying the general election ran from 29 October to 5 November, while its key dates say voting began on 28 October. The certified report gives the controlling exact interval: 28 October 2025 at 12:30 through 5 November at 13:00, America/Detroit. A members-meeting presentation dated 28 October says voting would open after the candidate forum and presents the package under Elections, Board Candidate Forum and Community Platforms. The slides corroborate the earlier date and the three-part framing; the certified report controls the exact time, wording and counts.
Those counts make the denominator structure fully visible. The report gives 725 electors and 184 ballots, so 184 / 725 is about 25.38% overall ballot participation. On the amendment question, 165 voted yes and four voted no: 165 + 4 = 169 non-abstaining responses. With 15 abstentions, 169 + 15 = 184, closing the ballot total.
The published 97.6% approval is 165 / 169, approval among non-abstaining amendment responses. Opposition is 4 / 169, about 2.4%. Abstentions are 15 / 184, about 8.2% of ballots. The response share is 169 / 184, about 91.85% of ballots. Finally, 165 / 725 is about 22.76%: yes votes as a share of eligible members. That last figure is not NANOG's published approval denominator, not a legal threshold, and not proof that adoption failed. It measures a different relationship.
The proposal, certified aggregate and current text close a 2025 proposal/result/final-text chain. They establish package adoption and textual effect. Yet precise arithmetic cannot reveal how support was distributed across 10.3.3, 8.1.1 and the 8.2 group. The 2025 record shows exactly what better reporting can repair—the meaning of the aggregate—and exactly what it cannot repair—the preference detail the ballot never asked for.
What one bit preserves, and what it erases
A binary package choice preserves a consequential fact: whether the electorate accepted or rejected the package presented. Where the process results in adoption, the package can acquire organisational effect as a unit. That is neither trivial nor merely symbolic. In both focal cases, NANOG disclosed the changes, put them to members and then incorporated the resulting provisions into its bylaws. Those facts strongly contradict any suggestion that the changes were hidden or imposed without a vote.
The same choice erases proposition-level distinctions at the point of measurement. In 2018, the ballot could not record separate answers for the vacancy procedure, membership discipline, committee oversight, staff administration, election notice and nondiscrimination language. In 2025, it could not distinguish the tabulation provision from the candidate-forum and communication-platform provisions. This conclusion requires no theory about voter motives. It follows directly from the number of controls offered.
High approval remains important evidence. NANOG reported 93 per cent package approval in 2018; in 2025, 165 of 169 non-abstaining amendment responses were yes. Those are strong package results on their stated denominators. But neither percentage can be distributed across clauses. Approval of a vector is not a separate measurement of every component within it.
This is an informational finding, not a Delaware-law opinion. The documented effect is that the disclosed text was adopted under NANOG's reported process. Whether either package satisfied every applicable legal requirement is beyond the public record used here. Governance legitimacy, certification practice, legal effect and a wider narrative about the technical community are separate questions; none can be read automatically from a turnout ratio.
For the same reason, eligible-electorate ratios should calibrate rather than invalidate. About 27.75 per cent is migrated-record overall election participation in 2018, not amendment response or eligible-member support. About 22.76 per cent is 2025 yes votes divided by all eligible members, not the certified approval test. Neither permits nonparticipants to be counted as no votes. Naming the denominator prevents a descriptive statistic from becoming an invented threshold.
A package manifest, not 34 mandatory questions
The practical answer is not to atomise every comma. It is to make packaging itself inspectable. Before a vote, a package manifest could number each proposition, show the affected section, quote the before and after text and state the purpose. It could label an effect as editorial, conforming, procedural or substantive while making clear that such labels are descriptive rather than dispositive legal findings.
For each proposition, the manifest could identify the operating surface: which actor gains, loses or has clarified authority; which right, duty, office, procedure or deadline changes; and which cross-references follow. In 2018, that would have placed the expired transition, grammar, vacancy rules, member discipline, committee oversight, staff moderation authority, interim-election notice and nondiscrimination language on one readable map.
The essential feature would be a dependency and severability statement. If several provisions must share a question, the sponsor should explain why separate outcomes could create contradiction, incompleteness or material risk. If an item could stand alone but remains bundled for economy, the manifest should say so and explain the trade-off. If the evidence does not support a confident severability conclusion, the statement should preserve that uncertainty rather than present convenience as necessity.
The manifest could make that judgment compact rather than burdensome. A short entry for each proposition could mark its dependencies as required, recommended or none identified; name the provisions affected by a partial result; and state the sponsor's chosen ballot group. Readers would then be able to audit not only the text but the architecture of consent. The important innovation is not another layer of prose. It is a public connection between each change and the reason it shares a voting control with the others.
That approach preserves the strongest reasons for packages. Interdependent provisions can remain together. A complete redline retains context. Members need not face dozens of isolated questions. But the outer boundary becomes a reasoned decision: the package explains why these provisions share fate, rather than allowing their co-location to masquerade as dependency.
The result report should then publish eligible members, ballots, question responses, yes, no, abstention and blank or nonresponse where the system records them. It should identify the approval denominator and keep question totals separate from election-wide totals. The 2025 report demonstrates how much interpretive ambiguity those fields remove. The surviving 2018 percentage demonstrates what remains uncertain when question-level counts disappear.
A manifest should also identify the correction route: who may propose a targeted amendment, when another ballot can occur, and whether an objectionable provision can be repealed without reopening the entire instrument. A later remedy cannot restore the preferences lost in the first package vote. It can make the consequences of package failure or post-adoption concern more manageable.
This is a bounded editorial recommendation, not a claim that NANOG currently uses such a form or is legally obliged to do so. Nor does every substantive clause require an individual question. The aim is to bring the resolution of the explanation closer to the resolution of the choice: number the propositions, show the authority shifts, identify dependencies, explain inseparability and publish the denominator chain.
The authority of the package, and the silence within it
The 18 September Board sentence remains the most economical description of the 2018 design: members would choose two directors and give one yes-or-no answer on all the bylaw changes. The package was visible, mixed and consequential. Its adoption produced operative rules on membership, Board continuity, committees, staff and elections. Its one result did not produce 34 opinions.
What bundling gained was the possibility of coherence: an integrated, counsel-reviewed redline; less ballot and administrative burden; protection against contradictory combinations; and a practical whole-package choice. What it surrendered was observability inside the bundle. The result cannot tell us which provisions had independent support, which encountered resistance, how intense any clause-level preference was or what trade-offs occurred.
The smaller 2025 package confirms the point under much better measurement. Its eligible, ballot, response, yes, no and abstention counts can all be reconstructed, yet its one question still cannot distinguish a view of ranked-choice tabulation from a view of community-platform governance. Better arithmetic tells readers precisely what the aggregate means. It cannot make the aggregate answer a question the ballot did not pose.
The choice is not between one question and 34 as a matter of dogma. A sensible package may be both legitimate and efficient. It should also reveal its construction: which provisions are linked, which are merely travelling together, where authority moves and why a common fate is justified. One switch can move many disclosed parts. It cannot speak for each part separately.
Metadata
- SEO title: The Ballot Had One Switch: NANOG's 34-Part Bylaw Package
- SEO description: NANOG's 2018 bylaw vote paired detailed disclosure with one binary choice. The 2025 record shows what denominator reporting can clarify—and what bundling still conceals.
- Social title: Thirty-Four Bylaw Changes, One NANOG Switch
- Social description: A complete redline can show every proposed change while one yes-or-no package vote still loses clause-level preference. Here is the case for bundling—and the case for a manifest.
Image notes
- Alt text: One unlabelled two-position ballot lever moves a board fitted with 34 differently shaped bylaw tabs; behind it, a smaller mechanism groups five tabs into three clusters.
- Caption: One control can move many visible provisions while recording no separate choice for any individual part.
- Accessibility description: A synthetic editorial illustration viewed from a slight angle. In the foreground, a physical two-position ballot lever is mechanically connected to a wide board of 34 tabs in varied shapes. A smaller assembly behind it groups five tabs into three linked clusters. The scene contains no readable ballot wording, logo, historical document, person or glowing network imagery. Its mechanical layout shows one binary action controlling many separately visible changes.
- Synthetic provenance: This wholly synthetic editorial illustration explains bundled governance consent. It does not reproduce a historical NANOG ballot, meeting room, document, logo or real person.