soc.culture.singapore.moderated

Long-form reference · 6,797 words · about a 30-minute read

What a moderated sibling was

On Usenet the word moderated named a routing rule before it named a temperament. A newsgroup carried a flag; the flag meant one thing only, which was that no article would be accepted into the group unless it arrived carrying an Approved header. Everything else people mean by moderation — judgement, patience, taste, and the perennial argument about who should be exercising them — sat on top of that single bit of server configuration. soc.culture.singapore.moderated was the second room of a pair: the same subject as soc.culture.singapore, largely the same readership, and a door that only opened from the inside.

The convention that put the word in the name was only a convention. The Big Eight creation guidelines constrained the shape of a group name — each component made of lowercase ASCII letters, digits, plus or minus, containing at least one letter, and no more than twenty characters long — and required the whole line of name, tab stop and short description to fit inside seventy-nine columns. Nothing in them obliged a moderated group to advertise the fact in its name. news.announce.newgroups, the moderated group through which every piece of the paperwork described below travelled, says nothing about moderation in its own name at all, and soc.culture.estonia, removed from the network in the same batch as this group, was moderated too. What the trailing .moderated component did was tell a reader scanning a group list, without having to consult anything, which of two adjacent rooms was the filtered one.

The flag itself lived in two places, neither of them in the group's name. On each server it lived in the active file, the list of groups that server carries: four fields per line — group name, highest article number, lowest article number, and a status field, in which y means local postings and articles from peers are both allowed, n means no local postings and articles only from peers, and m means the group is moderated and all postings must be approved. In the network at large it lived in the control messages that told servers what to believe. RFC 1036 — Standard for Interchange of USENET Messages, by M. Horton and R. Adams, December 1987 — defined the syntax as newgroup <groupname> [moderated], adding that if the second argument is present the group should be created moderated instead of the default, and that the message should be ignored unless it carries an Approved line of its own. The rule survived into the standards that replaced it: RFC 5536, Netnews Article Format (K. Murchison, C. Lindsey and D. Kohn), which defines the Approved header field itself, and RFC 5537, Netnews Architecture and Protocols (R. Allbery and C. Lindsey), both published in November 2009. In RFC 5537 the newgroup control message still takes exactly one optional flag, and the standard says so in as many words: moderated is the only flag defined by the protocol.

The third place the flag lived was the periodic checkgroups control message, which carries a list of all the valid groups in a hierarchy with their descriptions and moderation status and asks servers to make their own lists match. A moderated group's entry in such a list ends, literally, with the string (Moderated) — a piece of prose that is also a machine-readable field. RFC 5537 requires moderated newsgroups to be marked by appending that case-sensitive text; the Big-8 Board's own drafting rules add that it goes at the end of the line for every moderated group whether or not the word appears in the name; and the hierarchy administrators' FAQ notes, with visible irritation, that the word cannot be translated into another language because news software looks for exactly those characters. The line for this group read soc.culture.singapore.moderated Singapore & Singaporeans. (Moderated), and it read that way in the request for discussion, in both calls for votes, in the published result, in the control message that created the group, and in the vacancy notice posted fifteen years later.

There was a real difference between a group created moderated and an unmoderated group converted afterwards, and it was administrative rather than philosophical. Creation was a single control message that servers either honoured or ignored. Conversion asked thousands of independently administered machines to change a setting on a group they already carried, and the failure modes are set out at length in the Board's own documentation: some servers honour one message and not the other, so that the group appears moderated to some readers, unmoderated to others and missing entirely to a third set, while articles posted on the unmoderated side propagate badly and never reach the servers that are filtering. Hierarchy administrators were advised to send a few duplicate control messages after any such change, because it is harder to make stick than a plain creation, and to expect frustrated users whose providers simply never get it right. The Big Eight creation guidelines in force in the early 2000s went further and declared a moratorium: proposals to convert an unmoderated newsgroup to a moderated one would not be accepted at all. The Board remains on record against such conversions, and its stated advice is the arrangement this page describes — leave the original group intact and create a companion moderated group beside it.

The paperwork that made one

Creating a Big Eight group was Usenet's most elaborate civic ritual, and the arithmetic of it — the discussion period, the neutral votetaker, the hundred-vote margin and the two-thirds supermajority — is set out on the soc.* hierarchy page and not repeated here. What concerns a moderated proposal is the extra paperwork it had to carry. Under the guidelines maintained by the moderation team of news.announce.newgroups, a valid Request for Discussion had to contain a rationale for the proposal, charters for all newsgroups that would be created or changed, and moderator information sections for every group proposed as moderated. The Board's current statement of the same requirement spells out what those sections are for: who the initial moderators will be, including at least one decipherable and usable contact address. A moderated group could not be voted into existence anonymously. Somebody's name and mailbox went into the public record before the first ballot was cast.

For this group that record survives in full. The archive of news.announce.newgroups maintained by the Internet Systems Consortium keeps one file per proposal, filed under the proposed group's name, and the file for soc.culture.singapore.moderated contains five articles spanning fifteen years. The first is the Request for Discussion, posted on 15 July 1996 by Kelvin Tan and crossposted, as the rules required, to news.announce.newgroups, news.groups and soc.culture.singapore, with pointers placed in soc.culture.asean, soc.culture.malaysia and the Thai, Filipino and Indonesian groups. Its stated rationale was congestion: traffic in the parent group had become, in the proposal's words, too congested and afflicted by a low signal-to-noise ratio, and a moderated, more focused newsgroup was wanted. The proposal was explicit that the parent would remain active and unmoderated afterwards, to cater to readers who elected to stay in the unmoderated forum.

The vote followed the standard machinery. A neutral votetaker from the Usenet Volunteer Votetakers, Dave Cornejo, posted the first Call for Votes on 16 September 1996 and an identical second copy on 27 September, in accordance with the rule that a second CFV appear near the middle of the twenty-one-day voting period. Voting closed at 23:59:59 UTC on 7 October. The result was posted on 8 October 1996 under the standard headline format, RESULT: soc.culture.singapore.moderated moderated passes 192:48, and it announced a five-day discussion period in which allegations of voting irregularities could be raised, after which the moderator of news.announce.newgroups would create the group. The guidelines of the period were pedantic about the timing — control messages went out at 10:30 in the morning, United States Eastern time, Monday to Thursday, ordinarily at the first such time at least 120 hours after the result — and listed, among the permitted causes of delay, the initial setup of moderation. The room had to have a working letterbox before the door could be hung.

It went out exactly on that schedule. The newgroup control message for soc.culture.singapore.moderated is dated 14:30 UTC on Monday 14 October 1996 — half past ten in the morning, Eastern time, six days after the result — signed with the news.announce.newgroups PGP key, and it was repeated a week later. Its body names the submission address, the administrative address and all six moderators, and reprints the newsgroups line. So the answer to the question this page exists to answer is, unusually for a group of this size, a documented yes: there is a surviving record of this group's own proposal, its vote, its creation and its removal, with dates, addresses and names throughout. That is worth stating plainly, because for most of the groups in this directory no such record has ever been located and the honest page has to say so.

What the charter actually said

The charter published in the RFD and reprinted verbatim in both calls for votes, in the result and in the 2011 vacancy notice listed ten permitted subject areas, in the flat enumerative style that Usenet charters favoured: culture, religion and society; history and literature; travel, tourism and geography; politics, regionalisation and internationalisation; economics and business; education and science; computer networking; sports and entertainment; humour and languages; and nation building, development and values. Discussion could revolve around those topics but was, the document said, not limited to them. This was not an announcements group. It was the whole subject of the parent group, proposed again with a filter attached.

The off-topic list was shorter and more revealing, because it describes what the moderators expected to spend their time discarding: crossposts to many irrelevant newsgroups; “me too” posts, posts containing an unreasonable amount of quoted text, and duplicates; commercial and personal advertisements; binary files, even Singapore-related ones, which were to go to an appropriate alt.binaries.* group with a pointer left behind; obscene and pornographic posts; and posts consisting solely of insults or flames. Three of the six are problems of volume and format rather than objections to what a post says — the failure modes of an open discussion group in 1996, itemised by people who had been reading one.

On language the charter was careful. English was the preferred posting language; posts in Malay, and in non-ASCII Chinese and Tamil, were acceptable; anything in another language, alphabet or non-ASCII coding scheme was subject to consideration by the panel of moderators. What it cost to move those scripts through a seven-bit news feed in the middle 1990s is a story the parent group's page tells at length, and it is not retold here.

The section on the moderators' own conduct is short enough to quote, and is the part of the document that reads least like boilerplate:

The role of the moderators will be to ensure to the best of their ability and in good faith that a high signal-to-noise ratio of soc.culture.singapore.moderated is maintained and raised. The moderators will not edit any posts for spelling or grammar or whatever (unless the original poster requests for it). The moderators will not be held responsible for the contents of any submission posted. Submissions denied posting can be appealed and the appealed decision will be at the discretion of the moderator. Posts that have been denied to soc.culture.singapore.moderated can always be posted to soc.culture.singapore if the poster so pleases.

That last clause is the structural argument for the whole arrangement, stated by the people proposing it. A moderated sibling did not remove anything from the network; it added a filtered view of a conversation that continued, unfiltered, one line away in the group list. Rejection cost a poster the moderated audience and nothing else. Whether that is a satisfactory answer to the objection that moderation concentrates editorial power in very few hands is not for a directory page to say; it is, however, the answer this proposal gave, and it was given in advance and in public.

Six moderators were named, each with a paragraph of biography: Kelvin Tan, a Singaporean doctor with an interest in medical informatics; Tan Chong Kee, then a graduate student at Stanford; Wynthia Goh, working for a multimedia company in Singapore and active in soc.culture.singapore since late 1994; Harish Pillay, who is credited in his own paragraph with having created soc.culture.asean in 1989; Edmund Chia, a graduate student at the University of Tasmania; and Alvin Jiang, a school student in Singapore. Between them the six were resident in Singapore, Australia and the United States. The proposal added that volunteers would be recruited to join the panel if the need arose; whether any ever were is not recorded anywhere this research has found.

The electorate, counted

Because votetakers published their full ballot lists, the vote is one of the few numbers about this group that can be checked rather than guessed. The result records 192 yes votes and 48 no votes — 240 valid ballots — and 24 invalid ones. The invalid list is itemised with reasons, and the reasons are almost uniformly clerical: twenty-two of the twenty-four were rejected for containing no vote statement, one for an ineligible anonymous address, one for having no address at all. Voters had to return a specific form of words in the body of a message, and about one ballot in eleven of those received failed to.

The published list also records where the ballots came from, since each is printed with its address. Counting the domains in it puts a little over half of the valid votes — roughly 122 of 240 — at addresses in the .sg country domain, most of them at the commercial providers of the period, with the remainder spread across generic .com, .edu and .net addresses and the country domains of Australia, Canada, Britain, Japan and a scattering of European ones. The two sides were differently distributed: about three in five of the yes votes came from .sg addresses, against about one in five of the no votes. None of this measures readership, which nobody counted, and a vote-by-mail poll advertised in a handful of newsgroups is not a survey of anything. It does establish that a self-selected electorate of a few hundred people, about half of them writing from inside the country, turned out to settle the question, and that they settled it four to one.

For scale, the parent group's own creation vote, whose result was posted on 6 January 1993, passed 255 to 46 out of 326 ballots received, 25 of which were not counted. Nearly four years later the moderated proposal drew 264 ballots and 240 valid ones: an electorate of the same order, about a fifth smaller.

From newsreader to mailbox: the submission path

The mechanism that made moderation work was a diversion in the posting software, and it is the part of Usenet that modern readers most often get backwards. A poster to a moderated group did not post. The server their newsreader talked to recognised the group as moderated and mailed the article to a submission address instead of injecting it into the network. RFC 5537 states the requirement flatly: an injecting agent must forward the proto-article to the moderator of the leftmost moderated group listed in the Newsgroups header, customarily by e-mail, and if that is not possible must reject it outright. At the far end, a serving agent must keep a list of the moderated groups it accepts articles for, so that it can reject unapproved articles posted to them. Two independent checks, one at each end of the pipe, both looking for the same header.

Which left the question of what address to mail it to, and here Usenet did something characteristically pragmatic. Servers kept a moderators file mapping group-name patterns to submission addresses, in which a pattern could carry a substitution: the token %s was replaced by the name of the newsgroup with every period changed to a dash. The final catch-all line in the sample file shipped with INN, the standard news server software of the period, reads *:%[email protected], so a post to a moderated group nobody had configured locally went to the group name in dashed form at moderators.isc.org, where relays operated on the network's behalf forwarded it to whoever was actually reading. RFC 5537 records the same convention as the common method. The periods became dashes for a reason long since obsolete — periods in the local part of an address were not always handled correctly by the mailers of the day — and the documentation is candid that so much now depends on the transformation that it cannot easily be changed. Additions and changes to that central alias list were, and are, made by hand on request to a maintenance address at the consortium, with the warning that there is often a backlog.

This group did not use the central relay. Both calls for votes, the result and the creation control message publish its own addresses: articles went to [email protected], and administrative correspondence to [email protected]. Sintercom, the Singapore Internet Community, was a discussion site launched in 1994 by Tan Chong Kee — one of the six moderators — and its published list of former editors includes three of the six names on the moderation panel. The moderated newsgroup and the website were, in personnel and in plumbing, the same project reached from two directions.

What the moderator did on receipt is specified with more precision than most people expect. The article arrives as e-mail, either encapsulated whole or as a message carrying the news header fields directly; RFC 5537 tells moderators to be prepared for both. If it is accepted, the moderator adds an Approved header field identifying themselves and passes the article to an injecting agent. There are, the standard notes, no protocol restrictions whatever on the criteria used for accepting or rejecting a message, or on what a moderator may change in it; recommended best practice is to make the minimal required changes, with a warning that modifications may invalidate signatures created by the poster. Moderators are encouraged to retain the original Message-ID unless the article was significantly changed, and the original Date unless it appears stale — the standard's threshold is seventy-two hours or more in the past, and it names delays in the moderation process as a reason the date might be. Any Injection-Date, Injection-Info or Xref header fields already present must be removed. If the article was crossposted to a second moderated group, the moderator forwards it onward rather than posting it, and whichever moderator finally attaches the Approved header assumes responsibility for asserting that all of them approved it. The mechanics of the Approved header itself, and its place in the wider history of moderation, belong to the soc.* hierarchy page.

The moderator's working day

None of the above is the job. The job was reading everything, every day, unpaid, including precisely the material the group existed to keep out — which is the part of the arrangement the standards documents cannot describe, because it is not a protocol requirement but a habit of attention. A moderated group of any size meant one or more volunteers opening a mailbox each morning, reading submissions that had accumulated overnight from readers in several time zones, deciding on each, and injecting the survivors. Nothing about the mechanism degrades gracefully under fatigue. A moderator who reads half the queue publishes half the group.

The first cost was latency. An article written at midnight in one time zone reached a moderator who might be asleep in another, joined a queue, and appeared some hours or some days later. The seventy-two-hour staleness threshold in the standard exists because the standard expects exactly that: it tells moderators they may substitute the current date when the original has gone stale for reasons they understand, and gives delays in the moderation process as its example. The consequence for the texture of conversation follows from the arithmetic. A moderated group cannot interrupt itself. In an open group a misstatement is contradicted within the hour and the thread branches in three directions before lunch; in a moderated group the correction and the original may appear in the same batch, or the correction may arrive after the argument it was answering has moved on. Threads advance in shifts rather than in exchanges, and posters learn to write self-contained articles, because they cannot rely on the previous one having landed. The result reads less like a conversation and more like correspondence.

The second cost was concentration of effort in a place with no redundancy. Panel moderation, which this group adopted from the start with six names and an open invitation for more, was the standard mitigation: submissions could be distributed among several readers, holidays could be covered, and no single person had to be reachable at all hours. Robomoderation software later formalised the pattern, assigning each incoming article that needed review to a randomly chosen member of the team. But panels shrink quietly. Nobody resigns from a volunteer moderation team by press release; addresses stop working, jobs change, the queue is handled by whoever is left, and then by nobody.

Which is the standard failure mode, and it is worth naming precisely because it is the opposite of how an unmoderated group dies. The Big-8 Management Board's own account of the problem puts it in one line: an unmoderated abandoned group can be revived by dedicated users, whereas a moderated group without a moderator is dead. An abandoned open group fills with noise; an abandoned moderated group goes silent, because nothing can appear. Posts sent to the submission address bounce, or, in the Board's phrase, vanish into nirvana, and the poster is left with no explanation at all. From the outside the two states — a moderated group with nothing to say and a moderated group with nobody listening — are indistinguishable. The Big Eight eventually built an institution specifically to tell them apart, and it was used on this group.

Forgery, and the machines that answered it

The whole apparatus rested on a header that anyone could type. Nothing in RFC 1036 authenticated an Approved line; a server presented with an article bearing one had no way to know whether a moderator had attached it. RFC 5537, twenty-two years later, states the problem without embarrassment: a malicious poster may add an Approved header field to bypass the moderation process of a moderated newsgroup, injecting agents should verify that messages approved for a moderated group are actually being injected by the moderator, and there is at present no standardised method of authenticating approval of messages to moderated groups. The same document says the same thing about control messages, and the hierarchy administrators' FAQ says it more bluntly still: originally, control messages were authenticated only by the easily forged address of the sender, which worked while Usenet was small and broke down badly as it grew.

Two unstandardised systems filled the gap, and RFC 5537 cites both by name as being in common use. For control messages there was pgpverify, documented by David Lawrence in a specification titled Signing Control Messages dated August 2001 and distributed with a collection of hierarchy public keys, so that an administrator could configure their server to honour a newgroup message for a hierarchy only if it carried a valid signature from that hierarchy's key. For moderated articles there was PGP Moose, written by Greg Rose and cited in the same reference list under a specification dated November 1998. Its aim, in the description on the project's own page, was to monitor the news postings of moderators of Usenet newsgroups and to automatically cancel forged messages purporting to be approved: an approving moderator's software signs the article, and a second process watching the group verifies the signature on everything that appears and issues a cancel for anything that fails. Forgery was not prevented so much as made short-lived and conspicuous. In practice, as Board correspondence recorded much later, signed approvals never became common; most news servers simply refused Approved headers from anyone but trusted users, and that turned out to be enough for most groups.

Alongside the signatures came robomoderation — programs that sat in front of the human panel and did the mechanical part of the job. The best known on the Big Eight was STUMP, the Secure Team-based Usenet Moderation Program, written by Igor Chudov. It received submissions; rejected nonconforming ones automatically and without human intervention, megacrossposts and encoded binaries among them; maintained a list of pre-approved posters whose articles were presumed good and needed no review; distributed the remainder among multiple human moderators chosen at random for each message; and secured the exchange between the humans and the robot with PGP encryption and signatures, so that instructions to the robomoderator could not themselves be spoofed. In 2020 its author told the Big-8 Management Board he was content either to remain maintainer or to hand it over; Board members were subsequently appointed maintainers of STUMP and its web front end, and the code was moved to a public source-hosting service.

Both the danger and the answer can be watched happening to this group. The consortium keeps a second archive alongside the proposal files, holding every control message that ever named a given group, and the file for soc.culture.singapore.moderated contains ninety-seven of them. Two are the genuine creations of October 1996. Sixty-six were posted between 26 August and 18 November 1998, each purporting to come from David C Lawrence at the consortium, each carrying an Organization header reading The Cabal, each ending with the word Meow, and each announcing that soc.culture.singapore.moderated was an unmoderated newsgroup which had passed its creation vote by some figure the generator had invented — 252:45, 223:38, 295:0 and sixty-odd others, none of them the real 192:48. Every one of the sixty-six omitted the moderated flag, so a server that honoured them would have unlocked the door. Fifty of them carried an X-PGP-Sig header whose contents are not a valid signature. Interleaved with the forgeries and continuing to the end of November are twenty-four genuine, properly signed control messages from the same address, each reasserting the flag and reprinting the charter. The advice about sending duplicate control messages when a group's moderation status is in dispute was not theoretical.

The forgeries bear the marks of the Meow Wars, the long-running Usenet flame war whose participants signed their posts with the word, and whose floods of forged control messages are credited with pushing administrators towards digital signatures. This group's own creation message had carried a PGP signature since 1996, so an administrator running pgpverify would have discarded all sixty-six. A single unsigned rmgroup message in November 2001 asked servers to remove the group as bogus; nothing came of it, and the group stayed in the canonical list for another decade. By the time it was finally removed, the machinery for authenticating approvals was older than a good deal of the traffic it was authenticating: the vacancy notice posted about this group in June 2011 carries an X-Robomod header naming STUMP and its author, and two X-Auth headers containing PGP Moose version 2.0 signatures, one for news.announce.newgroups and one for news.groups.proposals.

Spam, and the group that could not be spammed

The strongest structural argument for a filtered sibling has a documented history behind it. Bulk posting to Usenet arrived as a recognised phenomenon in 1994, and the convention that settled after it defined spam by volume rather than by content — excessive multiple posting — because a network of that temperament could agree about quantity far more easily than about advertising. What followed, the cancel wars and the third-party cancelbots, the Breidbart Index that scored how widely a set of substantively identical articles had been spread, and the signed-notice schemes that deleted nothing and left the decision with each server, is set out on the alt.* hub page of this directory and is not repeated here. What belongs here is the one class of group for which none of that machinery was ever required.

Against all of that, a moderated group needed no defences, because it had no opening. Bulk posting works by injection: a script writes to many groups at once and the articles propagate. An article addressed to a moderated group is not injected anywhere. It is mailed to one address, where it sits until a person reads it, and if that person does nothing the article never exists as far as the network is concerned. A spammer targeting a moderated group is not spamming a newsgroup; they are sending a stranger an unsolicited e-mail, one copy, and the entire cost of the transaction lands in one mailbox. There was nothing for a Breidbart Index to score and nothing for a cancelbot to cancel, because nothing had appeared. Immunity to bulk posting was not a policy of moderated groups, nor a credit to their moderators; it was a property of the plumbing.

The corollary is that the immunity was bought entirely with someone's attention. The spam did not stop arriving; it stopped arriving in public. Every advertisement, every cross-posted provocation, every script-generated tract was read, or at least glanced at, by a volunteer before being discarded. This group's charter names commercial advertising, binaries and mass crossposts as off-topic in the same breath as flames, and STUMP-style robomoderation existed precisely to spare the panel the most mechanical part of that reading. The bargain a moderated group offered its readers was simple and it was never free: your newsgroup is clean because somebody else's mailbox is not.

What the record shows after 1996, and what it does not

The archive is nearly silent for fifteen years, and then produces the documents that settle the question. On 12 June 2011 Steve Bonine, posting through news.announce.newgroups on behalf of the Big-8 Management Board, opened a Moderator Vacancy Investigation into soc.culture.singapore.moderated. The procedure exists for exactly the failure described above: a formal investigation, begun because a moderated newsgroup is not functioning and may have been abandoned by its moderators, which may end in the installation of a new moderator, in the removal of the group, or — an outcome the Board records itself as considering likely to cause more harm than good — in conversion to unmoderated status. The published procedure would presume a vacancy in any group inactive for more than twenty-four months, and notes that the formal notice, crossposted into the silent group itself, effectively overrides that group's moderation on a temporary basis in order to be delivered at all.

The notice states its evidence in one line: probe posts to the group resulted in bounces or no response. It also states, under the heading history of the group, that the group passed its vote on 8 October 1996 by 192 to 48 and was active from October 1996 until May 2001. That sentence is the Board's summary rather than a traffic measurement, and it is the only statement about the group's working life that this research has been able to establish from a documentary source.

One adjacency deserves recording without any inference attached to it. The submission address for the group was at sintercom.org. Sintercom's own published history records that its founder received a letter from the broadcasting regulator on 5 July 2001 requiring the site to register under the class licensing scheme, that he registered it, and that he announced its closure on 20 August 2001. The Board's date for the group falling quiet — May 2001 — precedes that correspondence. The two events are separated by weeks and connected by a domain name and three shared names on a masthead, and nothing in the surviving record establishes any causal relationship between them, in either direction. The regulatory framework of the period belongs to the parent group's page, which sets it out in detail; what belongs here is only the note that the address a poster had to write to in order to reach this group was hosted by a site that ceased to exist in the same year the group did, and that the record does not say whether that mattered.

The rest is the ending, and it is fully documented. The June investigation was one of twenty-two posted between 29 May and 17 June 2011. On 19 July the Board published a Request for Discussion to remove twenty-three moderated newsgroups together — the batch it called the Great Downsizing 2011/2 — with soc.culture.singapore.moderated among them and a single rationale covering all of them: the groups had had no on-topic traffic for years, and the investigations had precipitated no interest in retaining them. Because of the size of the list the proposal was not crossposted to the groups it named; readers of news.announce.newgroups were asked to spread the message themselves. A Last Call for Comments followed on 16 August, opening the standard five-day comment period, and the result was posted on 30 August 2011: nine board members had voted to remove every group on the list. The rmgroup control messages went out from the consortium's group-admin address, PGP-signed, on 3, 4 and 10 September and again on 3 October 2011. In the namespace, the group had lasted fourteen years and eleven months. The present state of the list bears this out: the newsgroup list the Internet Systems Consortium maintains by processing control messages, which now runs to about forty-five thousand groups, roughly five thousand eight hundred of them flagged moderated, still carries soc.culture.singapore as an ordinary unmoderated group and does not list soc.culture.singapore.moderated at all.

What is not known should be said in full, because it is most of what a reader might want. No count of the group's readers or of its articles has been located. Nothing establishes which of the six named moderators were still serving at any given date, whether the panel ever recruited the volunteers the proposal anticipated, or what the group's traffic looked like in any year; the Board's own probe posts in 2011 produced bounces or nothing, which is evidence about 2011 and about nothing earlier. No thread, no article and no editorial decision from within the group is cited anywhere in the material behind this page; the charter is quoted here because it was published as part of a formal proposal, not because anything in this research has read the group. And the address itself survives for the reason the rest of this directory survives: it appears in the era's group lists and in the subscription roster of the mail-to-news gateway these pages document. That is a record of the group having been carried, which is not a record of what was in it.

Afterlife

Moderated groups aged differently from their open neighbours. The general decline of Usenet through the 2000s — readers leaving for the web, service providers dropping their news feeds, the traffic that remained shifting decisively towards binaries — thinned discussion groups everywhere; the culture branch as a whole emptied rather than died. But an open group with two readers is still an open group, whereas a moderated group with no moderator is a closed door. The Big-8 Management Board's response was procedural: the Moderator Vacancy Investigation, run in public through news.announce.newgroups with discussion in news.groups.proposals, a moderated group created in November 2006 for exactly this kind of business. The scale of the problem had been measured before the procedure was written: a note by Jim Riley dated November 2006, preserved in the Board's own account of the procedure, counted 131 moderated groups then inactive for at least twenty-four months, and the same page reckons that roughly ten to twenty moderated groups fall inactive each year. The investigations of 2011, this group's among them, worked through that backlog in batches.

Little of the branch's moderated furniture survived it. Of the tens of thousands of groups in the current canonical list, just two under soc.culture.* still carry the moderated flag: soc.culture.japan.moderated and soc.culture.russian.moderated. What survives instead is paperwork, and for moderated groups the paperwork is unusually good. Because every creation, every vote, every forged control message and every vacancy notice had to pass through, or be answered by, one moderated announcement group and one archive of control traffic, those archives constitute a near-complete administrative history of the Big Eight namespace, preserved as flat files and still served today. It is why a group with no located traffic archive can nonetheless be described here with dates, names, addresses and a vote tally. The proceedings survived better than the proceedings' subject.

The conversational archives are a different matter and are, for moderated groups, no better than for anyone else. Deja News began archiving Usenet in March 1995, sixteen months before this group's proposal; Google acquired the archive and Google Groups became operational in February 2001, supplemented by the end of that year with donated collections reaching back to 11 May 1981; and in February 2024 Google stopped supporting the posting or viewing of new Usenet content, leaving the existing archive in place. Coverage of any particular group in that corpus varies enormously, and the migration of the Deja archive is known to have lost older material during the relocation of its servers. A reader looking for what was actually said in a small moderated group in 1998 is dependent on whatever the archive happens to hold, which for many groups is little and for some is nothing.

Scope and limits

This article draws on four kinds of source and no others. The first is the standards record: RFC 1036 of December 1987 and its 2009 successors RFC 5536 and RFC 5537, which supply the definitions of the Approved header field, the newgroup and checkgroups control messages, the duties of injecting agents, serving agents and moderators, and the standing admission that approval was never authenticated. The second is the operational documentation of the Big Eight and of standard news server software: the creation guidelines posted by the moderation team of news.announce.newgroups, the Big-8 Management Board's published policies, minutes and vacancy-investigation procedure, the hierarchy administrators' FAQ, and the active file and moderators file formats. The third is the group's own file in the news.announce.newgroups archive — request for discussion, two calls for votes, result with the full ballot list, and the 2011 vacancy notice. The fourth is the archive of control messages naming this group, which supplies the creation and removal dates and the 1998 forgeries. Between them the third and fourth are the source of every date, name, address and number stated about this group in particular.

What that leaves out is the group itself. Nothing here rests on reading its articles, and no claim is made about what was posted, who posted it, how much of it there was, or how the panel exercised the discretion its charter gave it. The counts of ballot addresses given above are counts of a published list, offered as a description of an electorate and not as a proxy for a readership; the yes list yields 191 parseable addresses against 192 recorded votes, which is why the figures are given as approximations. Where the record is silent — on the group's traffic, on its later moderators, on the reason it fell quiet — this page says so rather than reasoning towards a plausible answer. Singapore itself, its connectivity, its regulation, its languages and its diaspora, belongs to the parent group's page; the hub-level history of Usenet moderation belongs to the soc.* hierarchy page. What this page owns is the procedure: how a room on Usenet acquired a door, who was named as holding the key, who tried to pick the lock, and what the surviving paperwork records about all three.