Summary

  • On 19 February 2025, NANOG called Discord a new official community channel. On 20 February, its Mailman migration notice promised to import list archives and preserve the old static archive for historical links. The public Discord announcement carried no parallel record-continuity promise.
  • The evidence does not show that Discord replaced the mailing list. NANOG’s public archive reports 35 discussions in both February and April 2025, while later meeting instructions continue to distinguish attendee email, live webcast chat and Discord.
  • Discord’s account, permission, edit and deletion mechanics make it useful for fast exchange but limited public evidence, by themselves, as a canonical public record. Those mechanics describe risk, not proof that NANOG lost messages or lacks private backups.
  • NANOG does not need to archive every conversation. A proportionate bridge would deposit formal decisions, corrections, reusable technical findings, authoritative links and named owners into a stable public surface, while leaving social, personal and safety-sensitive exchange private.

Two notices, one missing promise

The most revealing evidence is not a leak, a deleted message or a failed server. It is a pair of ordinary notices in NANOG’s own public mailing-list archive.

On 19 February 2025, NANOG’s newsletter announced the “New Official NANOG Discord.” The message said NANOG 93 might be over but conversation need not stop. It advertised channels for conferences, NANOG committees, technology and affinity groups, plus voice, video and screen sharing. That is a coherent offer. A real-time space can keep introductions, questions and working relationships alive between meetings in ways that a public email thread may not.

The next day, NANOG published a scheduled-maintenance notice for its mailing lists on the same archive page. The Mailman 3 migration was scheduled across 28 February and 1 March. It would move posting addresses from mailman.nanog.org to lists.nanog.org, migrate users transparently, preserve posting aliases and change the standardized List-ID headers. Then it made the promise that matters here: the archives would be imported, while the existing static archive would remain available for historical links.

That is continuity language. It names the thing being moved, the address change, a fallback condition, the treatment of users, the treatment of archives and the fate of old links. It does not guarantee that every message is perfect or that every old URL will survive forever. It gives a future maintainer and a public reader a testable intention.

The Discord announcement does something different. It names uses and features. It does not say which discussions are records, whether any public subset will be exported, whether messages can be cited through stable public links, how corrections are preserved, what happens if NANOG changes platform, or which surface controls when a chat comment conflicts with a list message, presentation, policy page or committee notice. None of the bounded public NANOG sources reviewed for this article supplies that parallel statement.

This is not proof that NANOG has no internal Discord policy, backup, moderator handbook or vendor-exit plan. Public silence cannot reveal private practice. Nor does the one-day sequence prove the same staff or Board decision joined the two projects. The defensible finding is narrower: NANOG publicly described record continuity for one communication system and publicly described participation features for another. The asymmetry is visible without guessing at motive.

“Official” changed faster than the record around it

The word “official” arrived against a more complicated channel history. NANOG did not move directly from email to Discord.

In June 2022, NANOG promoted a Community Forum as a digital place for community building, mentorship and ongoing support. The announcement listed technical areas such as automation, DNS, IRRd, RPKI and security alongside social and identity-based affinity groups. NANOG’s 2023 attendee archive continued to advertise that forum. The NANOG 90 instructions in February 2024 and the NANOG 91 instructions in June 2024 did the same.

Those messages establish a promoted surface and a stated purpose. They do not prove how active it was, who controlled it, whether its posts were public, or when its institutional role changed. The reviewed record does not publish a Community Forum closing notice, a data export, a map of old groups into Discord, or a decision explaining a succession. It would be wrong to fill that gap by calling the forum dead or claiming its content was deleted. Its later disposition is one of the unknowns.

By the time NANOG 93 opened, Discord was visibly present but not yet described consistently. The NANOG 93 welcome called it an “unofficial NANOG Discord server.” The same message nevertheless presented it as the digital place for affinity groups and a way to discuss the conference and keep conversation going between meetings. It separately described the attendee mailing list and a webcast chat monitored by Program Committee members for general discussion, speaker questions and technical help.

The official designation followed after the meeting. Yet the public announcement does not state who approved it, which rules changed, whether earlier messages became institutional records, or whether the former Community Forum’s content or groups moved. Nor does it establish that server ownership or moderation changed. “Official” is therefore a verified label with an unverified operational perimeter. It means NANOG endorsed the channel. On this evidence, it does not establish that NANOG owns Discord’s infrastructure, can guarantee message persistence or has made every channel authoritative.

The difference matters because readers often give “official” too much evidentiary force. An official venue can host jokes, introductions, job leads, speculative troubleshooting and provisional opinions. A speaker’s comment in that venue is not automatically a NANOG decision. A committee channel is not automatically a committee minute. A visible reaction is not a vote. Official status identifies the host relationship; it does not turn every utterance into an institutional act.

The list did not disappear

A replacement story would be neat and wrong. NANOG’s own archive blocks it.

The February 2025 NANOG archive displays a count of 86 people and 35 discussions. The April 2025 archive displays 96 people and 35 discussions. Those numbers are not a clean measure of operator participation. A person included in the archive interface is not necessarily a unique network operator; automated reports contribute traffic; a discussion count says nothing about readership, importance or influence. It cannot be compared directly with an unknown Discord denominator.

But the counts establish the point needed here: the public list remained active after Discord became official. There is no evidence basis for writing an obituary.

Later meeting instructions reinforce coexistence. NANOG 94 communications told attendees that they were automatically subscribed to an attendee list for updates and discussion. It described webcast chat for session discussion, speaker questions and technical support, monitored by Program Committee members. It then described Discord as a place to connect, collaborate and join affinity groups between meetings. A NANOG 94 newcomer message pointed specifically to a “Conference Chat” section on Discord while also publishing email addresses for staff and committees.

By May 2026, NANOG 97 communications still presented attendee email, live chat and Discord side by side. That looks less like replacement than a layered communications system. Email handles durable asynchronous threads and announcements. Webcast chat handles questions tied to a live session. Discord handles lower-friction, between-meeting and affinity conversation. Direct email provides accountable contact routes. The layers may overlap, but NANOG’s own instructions give them different jobs.

Fragmentation, not extinction, is therefore the right institutional question. A useful conclusion might begin in Discord, be tested on the public list, become a presentation and later be corrected on a website. Or it might never leave the channel. Without a publication rule, an outsider cannot tell whether silence in the public archive means nothing happened, something happened informally, or something happened and was later recorded elsewhere. That uncertainty should limit claims in both directions. It cannot prove hidden decision-making. It also prevents the chat surface from serving as reproducible public evidence.

What real-time chat adds

A skeptical audit should not pretend that email is the ideal social technology. The prehistory in NANOG’s archive shows why new surfaces appeared.

In a March 2021 archived discussion, one contributor said Discord was not a replacement for mailing lists but had a place, especially among younger groups. A message in the same exchange captured the access paradox: Discord can lower the barrier for people who already use it and raise the barrier for those who do not.

Those are contributor views, not demographic research or NANOG policy. They are valuable because they prevent a one-sided account. Mailing lists demand familiarity with thread etiquette, quoting conventions, inbox filters and a public archive that can make a novice’s question permanently searchable. A real-time channel can feel closer to walking up to a table. It can support voice, screen sharing and rapid clarification. It can make small affinity groups more viable between conferences. It can create a place for a question that is useful but not substantial enough for a global mailing list.

NANOG’s later promotion illustrates that breadth. Its August 2025 announcement archive invited real-time conversation about automation, BGP, peering, RPKI and other technical areas and republished a selected Discord example. The September 2025 archive highlighted a job-postings channel. This is not merely an online conference room. It combines technical exchange, professional opportunity, affinity organization and social continuity.

Those functions have legitimate value. The operator community is not built from formal papers alone. A contact made in chat can shorten an outage. A screen-share can help a newcomer see a configuration error. A job channel can widen access to the profession. An affinity space can keep someone in a field where they otherwise feel isolated. None of those benefits needs to become a public transcript in order to be real.

That is why the record problem must be stated carefully. The goal is not to force every hallway into a hearing room. It is to recognize when the outcome of a hallway conversation changes status.

The trigger is reliance outside the channel

A durable deposit is warranted when an exchange changes what people beyond the room are expected to know or do. A formal decision by an authorized NANOG body belongs in that body’s normal public record or an official notice. The deposit need not reproduce the conversation; it should identify the decision, date, scope, owner and governing authority.

Corrections need the same treatment. If chat corrects an official claim, meeting instruction, technical recommendation or published resource, the correction belongs next to the original public material. Otherwise only people in the right channel at the right time receive the safer information.

Technical findings present a different threshold. A troubleshooting exchange may isolate a failure mode, workaround or measurement caveat. Before others are asked to rely on it, someone should restate the finding on a durable surface with enough context to test it: affected versions, conditions, limits and a responsible author. Chat can discover a result; publication makes it citable.

Status changes and institutional undertakings complete the set. If Discord is the only place to learn that a service moved, a deadline changed, an incident was resolved or a committee accepted a task, the official web, email or governing record should carry the update and its current status. Convenience should not make an account-gated channel the hidden control plane for access. A dated public entry also distinguishes an open commitment from one completed, revised or withdrawn.

The threshold is not “interesting.” It is “material outside the channel.” Social exchange, tentative hypotheses, private assistance and ordinary peer discussion can remain conversational. A message acquires record value when someone beyond the people in the channel is expected to rely on it, when it alters an institutional obligation, or when it corrects evidence already in public circulation.

Reliance can emerge after the original exchange. A suggestion may begin as one operator’s provisional answer, then be repeated by a moderator, incorporated into meeting guidance and treated by attendees as the supported procedure. The first message need not be preserved merely because it started the chain. The record obligation arises when an identifiable NANOG actor adopts the result for an institutional purpose. At that point the public item should distinguish what NANOG is stating from what people in the exchange explored, and it should state the limits of the adopted result.

That approach preserves authorship and uncertainty without asking a later reader to reconstruct authority from roles, reactions or screenshots.

The reverse transition matters too. A chat message does not remain controlling merely because it once attracted attention. If a workaround is superseded, a deadline changes again or a committee withdraws a commitment, the durable surface should show the new status and point back to the earlier entry. The useful continuity is not an immutable transcript of every step. It is an intelligible chain in which a reader can identify the current statement, the responsible body and the reason an earlier statement no longer governs. This turns correction into part of the evidence rather than an unexplained disappearance.

A simple publication test follows. Would NANOG staff, a committee, a speaker or a meeting attendee reasonably direct someone outside the Discord channel to act on the result? If the answer is yes, that person should be able to supply a public link maintained on the appropriate NANOG surface. If the answer is no, the exchange can retain the informality that made chat useful. The test concerns institutional reliance, not popularity, technical sophistication or the number of reactions a message received.

Why Discord itself cannot supply the whole answer

Discord offers history and search within the product. That does not make it a public institutional archive.

Discord’s privacy policy says account creation involves credentials and contact or age-related information. It says users can edit or delete their own messages while they retain access, and people with appropriate permissions can manage servers, channels or other content. Discord’s developer documentation for messages makes access conditional on channel visibility and message-history permissions. Its search and retrieval interfaces are permissioned and paginated; deletion is part of the platform’s ordinary mechanics.

Those features are not defects in themselves. Permissions protect private groups. Deletion can be necessary for privacy, safety and error correction. Pagination is normal API design. The problem arises only if an institution silently relies on this user- and permission-dependent layer as its canonical record.

Discord also permits a user to request a data package. Its data-package documentation says the messages folder contains messages sent by that user, divided by conversation or channel. It also says a manually deleted message is no longer stored and will not appear in the package. A personal export is useful for an individual’s access rights. It is not a complete institutional export: it does not, by itself, reconstruct every user’s contributions, channel permissions, surrounding context, edits, moderation decisions or the public status of a claim.

The distinction is important. A platform may retain messages for product operation and still fail the requirements of a public archive. Public evidence needs more than storage. It needs a stable address, intelligible scope, provenance, correction rules, durable custody, access that does not depend on belonging to the right channel and a migration path when the platform changes.

This article cannot determine NANOG’s actual Discord configuration. The public sources reviewed do not reveal server ownership, moderator roster, role hierarchy, retention settings, backup method, export schedule, deleted-message policy, appeal route or vendor-exit plan. They also do not reveal whether private documents answer every one of those questions. Those are unknowns, not allegations.

The mailing-list standard is useful precisely because it is modest

The mailing list is not a legislature, and its archive is not a vote. NANOG’s usage guidelines describe an open, community-moderated forum. They prohibit abusive and irrelevant material, product marketing and unauthorized forwarding of private correspondence. The Mailman information surface exposes subscription and list-owner controls, while public archive pages give messages and threads linkable addresses.

These properties do not grant the list authority over networks. A thread can contain error, repetition, vendor interest, social pressure or a small and unrepresentative set of voices. Public visibility does not turn subscribers into principals for North American operators. Archive counts do not create a denominator for regional consent.

What the list provides is narrower: reproducibility. A reader can point to what was said, when, in what thread, and can often see later replies. A future maintainer can migrate that record while preserving historical links, as NANOG promised in February 2025. That makes list discussion evidence of discussion. It does not make the discussion a binding act.

This is the right standard to carry into chat. The objective is not to make Discord a legislature either. It is to ensure that an institution does not ask readers to rely on an exchange they cannot independently locate.

The operating rule

NANOG could close the gap without building an expensive archive of everything. A lightweight rule could have six parts.

First, name the canonical surfaces. The general list might remain the public technical-discussion record. The website might control policies, event instructions and committee notices. Formal minutes or election records would control institutional acts. Discord would be explicitly conversational unless a message is deposited elsewhere.

Second, define triggers. A decision, correction, reusable technical finding, official status change or institutional commitment made in chat must be summarized on the relevant canonical surface. The summary should contain date, responsible person or body, scope, source context and status. It should not need a full transcript.

Third, give deposits stable identifiers. A short web post, list message or decision-ledger entry can be cited later. If it changes, the earlier version should remain identifiable and the correction should say what changed. This is the same practical concern NANOG recognized when it promised to preserve historical mailing-list links.

Fourth, separate public evidence from private material. Safety reports, personal data, mentorship, employment inquiries, access needs and confidential operational details should not be copied into a public archive merely because they passed through an official platform. A deposit can state the resulting public action without exposing the protected conversation.

Fifth, assign an owner. Committee chairs, staff channel stewards or designated moderators can be responsible for recognizing record-triggering outcomes. People in the channel should also have a simple way to ask, “Is this now official, and where will it be recorded?” The answer should be a link, not a social-status judgment.

Sixth, publish a platform-exit minimum. NANOG need not expose backups or security architecture. It can state whether it can export public-designated material, how long that subset is intended to remain available, what happens to links, and where a successor record would live. The promise can be as proportionate as the Mailman notice: enough for readers to know the archive will not simply vanish with an interface.

Privacy is part of the record design

Permanent searchable chat can make a professional community worse.

People try ideas in conversation before they can defend them in public. Newcomers ask questions they fear will look naïve. Operators share partial symptoms during an incident. Members of affinity groups discuss experiences that are not institutional evidence. A complete public transcript can collapse the boundary between learning and publishing. It can magnify old mistakes, expose personal information, assist harassment and impose a moderation burden far beyond the value of the record.

Deletion also has legitimate purposes. A person may reveal a phone number, customer detail or security-sensitive fact by mistake. A moderator may need to remove abuse. A complainant may need a private route. A job seeker may not want a years-long public trace. Treating every edit or deletion as evidence tampering would be both inaccurate and unsafe.

There is also a selection problem. If moderators decide which chat moments become public, they can shape the institutional memory. The August 2025 newsletter shows that NANOG already selected a Discord example for wider promotion. That is ordinary editorial work, not proof of manipulation. But a publicity excerpt and an accountability deposit serve different purposes. The former chooses something engaging. The latter must follow a known trigger and preserve material context even when the result is awkward.

There is a credible alternative view: NANOG already maintains public mailing lists, public presentation and video archives, and direct staff and committee contact routes. Discord may simply be the rapid social layer, not a missing archive. That account fits the evidence better than a claim of wholesale channel replacement. It still leaves one practical condition. If NANOG expects outsiders to rely on a decision, correction, reusable finding or commitment first expressed in chat, the result must appear on one of those durable surfaces.

The deposit should be written as a concise public record, with consent from the person quoted where attribution is not already official, rather than copied wholesale from chat.

What the public record still cannot answer

The limits of the evidence remain substantial.

The sources do not identify NANOG’s Discord server owner, every moderator or the authority attached to each role. They do not state the server’s configured message-history access, backup or export practice. They do not publish a deleted-message retention rule, a moderation appeal path, a decision rule for committee channels or a succession plan if NANOG leaves Discord. They do not explain what happened to the earlier Community Forum. They do not provide comparable counts for mailing-list and Discord use, so no substitution rate can be calculated.

They also do not show harm. No source in the set reviewed for this article proves that a NANOG decision, technical correction or incident record existed only on Discord. No verified incident in this record establishes that a NANOG technical finding disappeared, that a member was excluded by the account requirement, that a committee acted without a public record or that Discord caused mailing-list decline. A design gap can exist before a documented failure. Calling it a gap is more precise than inventing a casualty.

Nor does record design enlarge NANOG’s authority. NANOG may set rules for its meetings, lists, community spaces and corporate bodies. It does not thereby govern an individual’s employment, ASN, routing, number resources, other Discord servers or conduct outside a stated scope. Preserving a chat-originated technical finding makes the finding inspectable; it does not make NANOG a regulator.

Classifying what crosses the bridge

The bridge becomes easier to administer if NANOG describes levels of record rather than treating every channel as either public authority or disposable talk.

At the first level is private or social conversation. It may be valuable to the people involved, but nobody outside the exchange should be expected to rely on it. A welcome, a recommendation for dinner, an informal introduction or a tentative troubleshooting hunch belongs here. The institution has no general reason to preserve it publicly.

The second level is discoverable community discussion. It may include a useful idea, link or question, but it has not been accepted by anyone with responsibility for a NANOG program. A Discord technical channel and a mailing-list thread can both sit at this level. The public list has an evidentiary advantage because outsiders can link to the thread, but visibility still does not make its content correct or official.

The third level is a reusable finding. Someone has supplied enough conditions, evidence and limits for others to test the claim. The right response is not to declare the entire chat authoritative. It is to ask the author, a channel steward or a relevant committee to publish the finding as a list message, presentation note, short article or maintained public document. The new item should identify its author and scope. If nobody is willing to stand behind the finding in a citable form, readers should treat it as a lead rather than a result.

The fourth level is an institutional statement. Staff or an authorized body has changed a deadline, corrected an instruction, announced a service condition, accepted a responsibility or recorded a decision. That statement belongs on the official web, notice, minute or decision-ledger surface regardless of where the first sentence appeared. A screenshot or forwarded chat excerpt is a poor substitute because it does not establish continuing status.

The fifth level is a controlling record. Bylaws, certified election results, approved policies and formal minutes have their own adoption and correction procedures. Chat should never silently amend them. If a discussion discovers that a controlling record is wrong or stale, the correction process should occur on the record’s own surface. This is the same discipline that keeps a mailing-list opinion from becoming a corporate act.

Such a ladder also protects Discord from being judged by the wrong standard. A social channel succeeds when people can connect safely and quickly. A canonical record succeeds when a later reader can reproduce what the institution committed to. The two systems can cooperate without pretending to be the same thing.

Testing the rule without inventing an incident

The policy can be tested through hypothetical but realistic operator-group situations. These are counterfactuals, not claims about events found in the NANOG Discord.

Suppose someone asks why an IPv6-only meeting network behaves differently from the legacy SSID. Several operators narrow the cause to a client implementation detail and identify a safe workaround. If the exchange only helps the people in the channel during that hour, it can remain chat. If staff begins directing all attendees to rely on the workaround, the status has changed. The meeting instruction or network page should carry the tested advice, its limits and a contact. A later correction should appear at the same address.

Suppose a committee member says a submission deadline has moved. Other users repeat the message. The underlying decision may be perfectly valid, but chat repetition cannot tell an outsider whether the member was speaking personally or for the committee. The cure is a short official notice. It protects applicants as much as the institution: everyone receives the same deadline, and staff does not have to adjudicate screenshots from different time zones.

Suppose an affinity group discusses a pattern of exclusion or a safety concern. Publishing the conversation could harm the people NANOG should protect. The durable record, if any, might be only an aggregate action: the responsible body received a concern, changed an accessibility measure or opened a review. Names, messages and sensitive detail remain private. Accountability does not require turning a protected report into public evidence about a person.

These examples locate the status transition. A conclusion crosses from conversation to record when institutional reliance begins. The boundary can be written without monitoring every channel for attractive content or treating routine deletion as suspicious.

Measure deposits, not audience size

It may be tempting to compare mailing-list traffic with Discord membership and declare which surface represents the community. The available evidence cannot support that calculation, and even a perfect count would answer the wrong question.

The February and April list pages expose people and discussions, but not the occupation, network role, geography, employer authorization or attention of every reader. A person who posts an automated routing report counts differently from a person who asks one operational question, yet both may appear in the same interface statistics. Discord membership would introduce other ambiguities: dormant accounts, people in many servers, multiple channels, lurkers and users whose employment or operational responsibility is unknown.

More importantly, representation is not produced by communication volume. A thousand chat members do not authorize NANOG to bind a network, just as a hundred list posters do not become an electorate for every operator in North America. Attendance, subscription and posting are stakeholder signals. Formal NANOG authority remains limited to the corporation, its members, its bodies and its own platforms.

The useful measures are more modest. How many official corrections first identified in chat reached the public material, and how quickly? How many reusable technical findings were deposited with conditions and limitations? How many institutional commitments have a visible owner and status? How many public links survive a platform migration? Those measures test the bridge without turning audience size into mandate.

They also keep the review away from private rooms. An auditor can check whether an accepted correction reached the relevant presentation, whether a committee commitment received a dated update, whether a deposited result names an institutional owner and whether its link still works after a defined interval. None of those checks requires reading private help requests or scoring individual people. The review covers NANOG’s work on the surfaces it controls.

Assign the handoff

A deposit rule fails if responsibility belongs to “the community.” Someone needs the verb.

For meeting webcast chat, the NANOG 93 and NANOG 94 instructions already identify Program Committee monitoring. Monitoring need not mean that committee members endorse every message. It does make the role a plausible first detector for session questions or corrections that should reach the speaker page or meeting notice.

For committee-related Discord channels, the committee chair or a named channel steward could identify formal commitments and route them to the committee’s existing record. For general technical channels, the author of a finding should normally publish it; a moderator can invite the step but should not convert another person’s provisional words into an endorsed technical conclusion. For staff announcements, the staff member or responsible program owner should update the canonical page.

The system also needs a challenge path. Anyone who sees others treating a chat statement as official should be able to ask for its canonical link. If none exists, the answer should say whether the statement is informal, under review or awaiting publication. That small discipline reduces the power of insider memory. The person with the most access to private channels no longer becomes the only interpreter of what NANOG “decided.”

This is not bureaucracy for its own sake. A deposit can be five sentences. The cost should track the consequence. A changed deadline needs a quick notice; a reusable technical claim needs evidence and limitations; a corporate decision needs the record already required by its governing process. The principle is consistent while the form remains proportionate.

Plan for the next platform

Discord will not be the last community tool. NANOG’s use of lists, a Community Forum, meeting chat and an official Discord as distinct or overlapping surfaces shows why a platform-neutral rule matters more than a platform-specific archive project.

Before adopting a successor, NANOG can ask a compact set of questions. Can public-designated material be exported in a documented format? Can stable identifiers survive an interface change? Can corrections point back to earlier text? Can access be provided without exposing protected channels? Who owns the institutional account and recovery credentials? What happens if the vendor suspends the service, changes terms or removes an export function? Which material must be deposited elsewhere before shutdown?

Those questions do not require NANOG to self-host every conversation. Vendor platforms often provide better usability, moderation and real-time features than a small association could build. Portability is not a demand for technological isolation. It is a refusal to let the only copy of an institutional commitment depend on one interface and one set of permissions.

The February 2025 Mailman notice already demonstrates the pattern: announce the window, describe addresses and headers that change, migrate users, import archives and preserve historical links. A future chat migration will differ technically, but its public promise can be equally legible.

The deposit experiment can remain small and reversible. NANOG could define a few categories, exclude social and safety-sensitive exchanges, and review the workload after a fixed period. If almost nothing crosses the threshold, that would support the view that Discord is mainly a fast social layer. If corrections or commitments do cross it, the bridge would have shown its value without turning conversation into an archive. A large server membership, high message count or popular reaction would still establish no mandate.

The public entry would identify the narrower act: a speaker corrected a published claim, a committee accepted a task, or NANOG chose to maintain a resource.

NANOG’s Mailman notice did not claim that email represented every operator. It told readers what would move, what would change and how history would remain reachable. Discord needs a promise proportionate to chat: consequential outcomes leave the room for the appropriate public record, while ordinary conversation and protected material do not. The institution can keep the speed of real-time exchange. It must make the boundary with institutional memory visible.

Sources