soc.answers

Long-form reference · 8,167 words · about a 36-minute read

A newsgroup that could contain nothing of its own

The clause that distinguishes soc.answers from its parent is one sentence long, and it was written into the call for votes of 19 November 1992 by the moderator of news.answers, Jonathan I. Kamens of MIT. It survives verbatim in the newgroup control message that created the group, which reproduced it under the heading The charter, culled from the call for votes:

The charters of these newsgroups will be identical to the charter of news.answers, with the addition that each of these newsgroups will contain only periodic informational postings that are also posted to other newsgroups in the same hierarchy (e.g., the “comp.answers” newsgroups would contain only postings that appear in at least one other “comp” newsgroup).

Read carefully, that is a charter for a newsgroup with no content of its own. Every article in soc.answers was, by rule, the second copy of an article already posted in some other soc.* group; a document that appeared only in soc.answers was out of charter, and so was a document from any other hierarchy. The group was not a place where anything happened. It was an index in the shape of a newsgroup: a single address at which the standing reference literature of a hierarchy of a couple of hundred discussion groups — 213 of them in the Internet Systems Consortium’s current active file — could be found, subscribed to, archived, expired on its own schedule and dropped by a site administrator as a unit.

That inversion is the whole point of the thing, and it is the reason soc.answers is worth a page of its own even though almost nothing was ever said in it. A discussion group is interesting because of its traffic. soc.answers was interesting because of its redundancy. The family it belonged to, and the institution behind that family, are described on the news.* hub, which owns the story of news.answers, the periodic posting and the archive that collected them; what follows is the soc.* member’s own record, the arithmetic of the family it belonged to, and the literature it carried.

The line in the newsgroups file

The description the group still carries is not editorial prose written for a directory. It is a line in a data file. The Internet Systems Consortium distributes a newsgroups file from which any news server can build its own group table, and in the copy consulted for this page the entry reads, with tabs between the two fields:

soc.answers Repository for periodic USENET articles. (Moderated)

The parenthesis at the end is not decoration and not an editor’s note. RFC 5537, the Netnews architecture standard published in November 2009, specifies the format of exactly this line and requires that the description “MUST NOT contain any occurrence of the string ‘(Moderated)’ within it” and that moderated newsgroups “MUST be marked by appending the case-sensitive text ‘ (Moderated)’ at the end”. The suffix is a machine-readable flag with a fixed spelling, and a description that happened to contain the word in its own right would corrupt it. The companion active file, which is what a server actually consults, states the same fact more tersely still: the entry for soc.answers ends in the single character m. The group is, in the files consulted on the day this page was written, still flagged moderated.

The description itself is older than the group. Ten entries in the current newsgroups file share the sentence “Repository for periodic USENET articles.” word for word — alt.answers, comp.answers, humanities.answers, ia.answers, misc.answers, news.answers, rec.answers, sci.answers, soc.answers and talk.answers — nine of them with the moderation marker and one, ia.answers, without. It was written for news.answers, and it can be dated: the newgroup control message that created news.answers on 23 September 1991 already carries the line news.answers Repository for periodic USENET articles. (Moderated), above a note that the group had passed its vote 337 to 19. From there it was copied into every control message that created a sibling. A reader who looks soc.answers up in a newsreader today is reading a sentence composed for a different newsgroup more than a year before soc.answers existed.

The group’s own paperwork, and where it is not

Big-Eight groups are unusually well documented, because the Internet Systems Consortium mirrors two separate archives: the news.announce.newgroups postings, filed one text file per proposed group, and the newgroup and rmgroup control messages actually seen on the network. Anyone checking soc.answers against the first of those will draw a blank. There is no file named soc.answers in the soc directory of the news.announce.newgroups archive; the listing steps from soc.adoption.parenting straight to soc.apathy.

The paperwork is not missing. It is filed elsewhere, and the reason is instructive. soc.answers was never proposed on its own. It was one line item in a single proposal to add six hierarchy-level companions to news.answers, made by the moderator of news.answers, and the archive files a proposal under the name the proponent gave it. The file is called news.answers-expansion, it sits in the news directory, it runs to some sixty-two kilobytes, and it contains the whole cycle: the request for discussion of 18 October 1992, the call for votes of 19 November, the second call and vote acknowledgement of 4 December, and the result of 6 January 1993. Every one of the four is signed by Kamens at athena.mit.edu and carries a double approval — from [email protected], then the moderator of news.announce.newgroups, and from [email protected].

Expansion is the right word for it, and the call for votes is emphatic on the point. news.answers was not being broken up. It would continue in its current form, and, in the proposal’s capitals, all periodic postings would continue to appear in it even when they were also crossposted to one of the new groups. What the six new groups added was a filter: a site or a reader could now take the reference literature of one hierarchy without taking the rest.

The documents also carry an auxiliary header line reading Archive-name: news-answers/administrivia. The paperwork that created soc.answers was itself an *.answers periodic posting, archived under the same scheme as the FAQs the group was built to carry. There are worse one-line summaries of the whole enterprise.

The RFD’s diagnosis of what had gone wrong with news.answers, and the full table of results for all six groups, belong to the news.* hub and are not repeated here. Two features of the ballot are worth recording on this page because they are specific to how soc.answers came to exist. The first is that it was a single six-way ballot. Voters mailed [email protected] with one line per group, in the form vote soc.answers MYVOTE, replacing the last word with yes, no or abstain; the last vote received for a given group was the one counted, and posted votes were void. For sites whose mailers could not resolve the address, the call for votes supplied three alternative spellings of it, two of them bang paths through mit-athena and uunet and one a percent route via athena.mit.edu.

The second is what the six columns show. Yes votes across the six groups ranged from 232 to 258 and no votes from 27 to 35 — a spread of twenty-six and eight respectively, on turnouts of well under three hundred. soc.answers recorded 240 in favour and 29 against. Whatever the electorate thought it was voting on, it was plainly not voting on the individual hierarchies: the six results are the same result six times, with noise. Kamens noted in the same posting that both calls for votes had misdescribed the closing date as a Saturday when 20 December 1992 was a Sunday, and that this had made no difference because no votes arrived on the 20th. He also confirmed that he would initially moderate all six groups himself, and asked for volunteers to divide the work.

12 January 1993, 12:42:13 GMT

The control archive holds the moment of creation to the second. On the morning of 12 January 1993 David C. Lawrence, posting as [email protected], issued six newgroup messages in nine minutes and twenty-two seconds: comp.answers at 12:33:48 GMT, misc.answers at 12:37:44, rec.answers at 12:39:12, sci.answers at 12:40:20, soc.answers at 12:42:13 and talk.answers at 12:43:10. soc.answers was the fifth of the six. The result had been posted on 6 January; the delay is the customary objection window, whose mechanics are described on the news.* hub.

The body of the message is short. It states the mandate — that soc.answers is a moderated newsgroup which passed its vote for creation by 240:29 as reported in news.announce.newgroups on 7 January 1993 — gives the line for the newsgroups file, quotes the charter, and disposes of the question of who was to run it in a single sentence:

For now the relevant addresses are the same as for news.answers.

The submission address was [email protected] and the moderator contact address [email protected], both attributed to Jonathan Kamens. soc.answers therefore began life without a moderator of its own, and there is no record in the control archive of it ever acquiring one. It was administered for its whole working life by a small collective team addressed at a single role account, which is why no individual name can honestly be attached to this group in particular.

A census of the answers groups

Nobody appears to have counted them, so here is the count. In the ISC newsgroups and active files consulted for this page, thirty-one newsgroups have answers as their final name component. Twenty-six are flagged moderated in the active file; five are not.

• The Big Eight, all of them: comp.answers, humanities.answers, misc.answers, news.answers, rec.answers, sci.answers, soc.answers, talk.answers. Every one of the eight hierarchies has an answers group, and every one of the eight is flagged moderated.
• Other global and national hierarchies: alt.answers, de.answers, uk.answers, pl.answers, cz.answers, cn.answers, han.answers, tw.bbs.answers, nctu.answers, chinese.newsgroups.answers, fido7.news.answers, fido7.ru.news.answers.
• Sub-hierarchy and site-level groups: alt.comp.answers, arkane.answers, bcs.answers, ca.answers, cuug.answers, demon.answers, free.answers, halcyon.answers, ia.answers, own.answers, and fiod7.other.news.answers — the last of which is a misspelling of fido7 that has been propagating in group lists long enough to have an entry of its own, carrying the description “No description.”

Several of those descriptions are worth reading as evidence of how far the form travelled. de.answers calls itself a Fundgruppe fuer regelmaessige deutschspr. USENET-Artikel, the coinage evidently playing on Fundgrube, a treasure trove, and it still carries [email protected] inside the description line itself; uk.answers is a “Repository for periodic UK specific USENET articles”; cz.answers offers “FAQ - casto kladene otazky a odpovedi”, frequently asked questions and answers; pl.answers promises “Odpowiedzi na pytania, o ktorych wam sie nie snilo”, answers to questions you never dreamed of; cuug.answers offers “Canadian Unix Users Group helpful documents”; ca.answers, more briskly, offers “All the answers for California”; and bcs.answers promises frequently asked questions about the “Bostom Computer Society”, a misspelling that has now outlived the organisation it misspells, the Boston Computer Society having wound up in 1996.

The five unmoderated ones are alt.comp.answers, bcs.answers, free.answers, ia.answers and the misspelt fido7 entry. That distribution is itself the finding: the answers idea was inseparable from moderation, because without a gate there is no way to guarantee that the group contains only approved periodic postings, and a repository that anyone can post to is simply another discussion group.

Equally telling is the list of hierarchies that never adopted the form. Among the larger national and regional trees in the same files, fr.*, it.*, es.*, nl.*, aus.*, no.*, dk.*, fj.* and japan.* have no answers group at all, and neither does relcom.*. Several of them are substantial: it.* carries 449 groups in the active file, japan.* 490, fj.* 412, dk.* 247, no.* 219, relcom.* 193, es.* 180, nl.* 177, aus.* 148 and fr.* 144. The English-language Big Eight, the German, Czech, Polish and British hierarchies and a cluster of East Asian ones took up the convention; the French, Italian, Spanish, Dutch, Australian, Scandinavian and Japanese ones did not. The reasons are not recorded anywhere this page could find, and it would be easy and wrong to invent them.

The family did not arrive in a tidy order

The control archive gives a first-sighting date for each of the surviving answers groups, and the sequence is untidier than the usual account suggests.

• news.answers — 23 September 1991, the original, created on a vote of 337:19.
• rec.aviation.answers — earliest newgroup message in the archive dated 1 October 1992, more than three months before the hierarchy-level groups. Its later removal notice dates the group itself to September 1992.
• comp, misc, rec, sci, soc and talk.answers — 12 January 1993, in one nine-minute run.
• alt.answers — 3 February 1993, issued not by the Big-Eight machinery but by Alan Greig at dct.ac.uk, alt.* having always run its own creation rules.
• de.answers — 7 December 1993, from subnet.sub.net rather than from any central authority.
• uk.answers — 5 March 1995, issued by Jon Thackray from a Demon address.
• humanities.answers — 25 July 1995, passing 164:22, about three months after humanities.misc was created on 17 April 1995 and turned the Big Seven into the Big Eight.

rec.aviation.answers is the interesting anomaly, and not only because it came first. It shows that the answers convention was applied at whatever level of the tree had enough documents to justify it: a single busy sub-hierarchy could have its own repository. It is also the only member of the family whose ending is documented in detail, which makes it useful evidence later in this page. It was removed by the Big-8 Management Board in March 2007.

Moderation that did not moderate

What moderation meant in the soc.* hierarchy generally — the Approved header, the volunteer with an inbox, the honour system — is set out on the soc.* hub. The *.answers groups worked differently, and the difference is the most easily missed fact about them. The moderators did not post anything. The introductory document that the team posted monthly to the answers groups says so in as many words, and flags the point as one that needs emphasis:

This needs to be emphasized, as it marks a very important difference between how the *.answers moderated newsgroups work and how most moderated Usenet newsgroups work: we will not actually post copies of postings which appear in *.answers. Instead, our job is to approve such postings (which are then cross-posted by their maintainers directly into the *.answers newsgroups) and to watch over the *.answers groups for problem postings.

A maintainer submitted a document once. If it was accepted, the moderators explained how to mark the article as approved in its header, and the maintainer then posted the document into the moderated groups directly, on their own schedule, for as long as they liked. There was no per-issue review. A FAQ approved in 1994 and posted monthly until 2001 passed a moderator’s desk exactly once.

Nor was the review editorial. The submission guidelines are explicit that the content was none of the moderators’ business: they “don’t cover the format or content of your article at all, as long as it’s periodically posted, human-readable information of some sort, and not overly commercial”, with unbiased discussions of commercial products welcome and advertisements not. What they did specify was headers. The *.answers moderators were registrars, not editors; they checked that a document was addressed correctly, named correctly for the archive, and unlikely to break anything, and then got out of the way. The single discretionary power the charter granted them was inherited from news.answers and stated in the first person: where there is an ambiguity, the moderator will decide whether or not a posting belongs in the newsgroup.

The team was small, collective and permanently behind. Correspondents were asked not to write to any individual moderator’s address, even the one who had dealt with them most recently, because it would only delay the processing of their submission; from February 1995 every submission and every piece of e-mail was automatically acknowledged, because the team “receive thousands of submissions, correspondence, and other email each month” and could not always answer. The copy of the guidelines posted in May 2004 was signed by David Alex Lamb and Nick Boalch, with Jonathan I. Kamens, Ping Huang and Pamela Greene listed as emeritus. Five names, across roughly thirteen years, for the registry of Usenet’s entire reference literature.

Why the order of the Newsgroups line mattered

The guidelines’ most pedantic-looking rule turns out to be load-bearing. A submission had to list its home newsgroups first, then the relevant hierarchy answers groups, with news.answers last of all; the checklist asks five separate questions about that one line. The reason is in the protocol. Under RFC 5537, when an article naming one or more moderated groups arrives without an Approved header, the injecting agent must forward it “to the moderator of the leftmost moderated group listed in the Newsgroups header field”. Ordering the line therefore decides which volunteer’s inbox an unapproved article lands in.

Put the home group first and a mis-posted article goes to that group’s own moderator, or straight onto the network if the group is unmoderated, rather than into the *.answers queue. Put news.answers last and an article crossposted to several answers groups is handled by the hierarchy moderator before the global one. The guidelines spell out the consequence for a document destined for both soc.answers and some other moderated group: the maintainer had to obtain approval from every moderator separately, was advised to get the others in first in case one of them refused, and had to submit to the *.answers team by e-mail rather than by posting, because posting would have routed the article to whichever moderated group happened to sit leftmost. RFC 5537 describes the same chain from the moderators’ side, where each in turn adds an indication of approval and forwards to the next unapproved group.

The rule was not always obeyed, including by the documents the moderators themselves archived. The copy of the soc.culture.jewish FAQ examined for this page carries Newsgroups: soc.culture.jewish,soc.culture.jewish.moderated,news.answers,soc.answers — news.answers third of four rather than last, with soc.answers behind it. Its Approved line names two authorities, [email protected] and an address for the moderated companion group, which is exactly the multiple-moderator case the guidelines describe.

The same logic governs the other required header. Every submission had to carry a Followup-To line, and that line was forbidden to contain any answers group. Replies to a FAQ therefore went to the home group or, if the maintainer preferred, to the maintainer — the value poster — and never into soc.answers. This is why the group is silent in a way that no ordinary moderated group is silent. It was not merely that few people wrote to it; the standard configuration of every document in it made writing to it impossible by accident.

The auxiliary header

Below the ordinary Usenet headers, separated by exactly one completely blank line and followed by at least one more, every *.answers posting carried a second block of headers of its own invention. This was the machinery that turned a stream of articles into a filesystem.

• Archive-name (required). The document’s path in the archive. Words were separated by hyphens and slashes marked levels of hierarchy, because, as the guidelines explain, the components between slashes became directory and subdirectory names when the posting was filed; each component was to be kept under fourteen characters, or at least to carry the important part in the first fourteen. Maintainers were asked to choose a name comprehensible to someone outside the field, describing the contents rather than the newsgroup, and were encouraged to cluster related documents under a shared first component.
• Posting-Frequency (optional but effectively compulsory). Free text, in any format a human reader would understand. If you did not supply it you had to tell the moderators anyway, so that the frequency could go into the List of Periodic Informational Postings.
• Last-modified, Version, URL, Copyright, Maintainer (all optional). The archive scripts and other software recognised these particular names and might handle them in special ways, so maintainers were asked to use the exact spellings rather than variations on the theme.
• Newsgroup-specific archive names. A line of the form Newsgroup-name-archive-name: overrode the general Archive-name when the document was filed under that particular group, so the same article could be stored as one path under the answers groups and another under its home group.

Two ordinary Usenet headers did the work of keeping the archive current, and their standards history is a small comedy. Expires has been in the specification since RFC 1036 of December 1987, where it is defined as a suggested expiration date, where users are discouraged from setting one unless the topic has a natural expiry, and where system software is told it “should almost never provide a default” one. FAQ maintainers used it for exactly the case the RFC had described, setting it far enough ahead that the current issue survived until its successor arrived. Supersedes, which replaces the previous copy outright, appears nowhere in RFC 1036 at all. It was long-established practice before it was standardised, arriving only in RFC 5536 of November 2009 — the document that obsoletes RFC 1036 — complete with a note that has the weary air of a long-running support ticket: “There is no ‘c’ in Supersedes.” The *.answers guidelines had been issuing the same warning for years, adding that most Usenet software would simply ignore a variant spelling.

Two further rules of thumb shaped the rhythm of the group. Maintainers were told that a document not posted at least once every three months, and not declared in advance as less frequent than that, might disappear from the rtfm.mit.edu archive, because the automatic cleanup scripts assumed silence meant the document was out of date. And they were asked to pick a random time of the week or month rather than the obvious date: if everyone posted on the first, the resulting flood would “overwhelm some smaller news sites and many readers of *.answers”. Being asked to randomise your posting date to protect the network is as good an illustration as any of the scale the system was designed for.

The posting robot

The guidelines are careful to say that many maintainers posted their files by hand with no problems at all. They then devote several pages to the packages that would do it for you, and it is the automated route that shaped the archive. Three pieces of software are named, all of which handled Expires, Supersedes, References and multi-part posting on a schedule.

• post_faq, written by Jonathan Kamens and requiring Perl, distributed from rtfm.mit.edu as a shell archive and, by the time of the later guidelines, maintained by the *.answers moderators themselves. It took an article with its static headers, added the ones that changed with each posting, and posted it.
• auto-faq, originally by Ian Kluft and later maintained by Paul W. Schleck under the GNU General Public License, offering more automation including the automatic building and insertion of all headers; version 3.3.1 is the one named in the guidelines.
• The FAQ server, a mail-to-news robot run by the moderation team at rtfm.mit.edu, recommended for maintainers whose sites had no Perl, or who could not conveniently install either package, or — the case that says most about the era — whose local news server would not let users post to a moderated newsgroup at all. You mailed it your document and your schedule and it posted for you.

Submission itself was automated too. The recommended route was the FAQ-checker at [email protected], which validated a posting against the guidelines, mailed back a message explaining what was wrong with it, and forwarded it to the humans only if it passed; articles it had okayed could be processed more quickly. It did not understand MIME, so submitters were told to turn off encoding and quoted-printable, avoid stray eight-bit characters such as accents and smart quotes, and stop their mailers wrapping long lines like the Newsgroups header. A maintainer who wanted a dry run could put the word ignore in the subject and get the diagnosis without the submission.

The consequence of all this automation is worth stating plainly, because it explains the shape of the archive that survives. Once a document had been approved and a robot pointed at it, it kept arriving. Maintainers moved on, changed employers or simply lost interest, and their FAQs went on appearing every month regardless, revised less and less often, until something broke or someone noticed. The genre’s greatest strength and its eventual failure mode were the same mechanism, and the specimen examined in the next section shows it clearly: a document whose last substantive change is dated October 1995 was still being posted, unaltered, in March 2004.

What a soc.* FAQ actually was

The specimen document that the *.answers moderators chose to illustrate their own rules was a soc.* one. The worked example at the top of the submission guidelines shows a posting headed Newsgroups: misc.foo,soc.culture.foo,misc.answers,soc.answers,news.answers with the subject line “[soc.culture.foo] Welcome - read this first!” and a summary describing the newsgroup and where to find more information. When the people who ran the system reached for a picture of a typical periodic posting, they drew a culture-group welcome document.

That is the genre in miniature. A soc.* FAQ was rarely a list of questions in any strict sense. It was a group’s standing paperwork: what the room is for, what belongs in it and what belongs next door, the transliteration or romanisation conventions the group has settled on, the arguments that have already been had and need not be restarted, the reading list, and the reference matter that a stranger would otherwise ask for. How that document functioned socially inside a culture group — as charter, etiquette manual and shield — is described on the soc.culture.* hub. What concerns this page is its physical form, and there the record is unusually good, because the documents themselves survive with their headers attached.

Take a specimen that can be checked line by line. Part one of the soc.culture.jewish FAQ, in a copy posted on 28 March 2004, is subtitled “Introduction to the FAQ and s.c.j Newsgroups (1/12)”. It carries Archive-name: judaism/FAQ/01-FAQ-intro, a newsgroup-specific archive name filing it separately under soc.culture.jewish, a Posting-Frequency of Monthly, a Followup-To of poster, two URL lines pointing at web copies, a Summary, and an Expires header set about six weeks ahead of the posting date. It was maintained by Daniel P. Faigin and, on the evidence of its own headers, posted not by him but by an automated mirror on his behalf. Its high-level index runs to twenty-one numbered sections distributed across the twelve parts, covering network and newsgroup information, who the contributors are, Torah and halachic authority, holidays, dietary law, Sabbath and holiday observance, women and marriage, worship, conversion and the question of who is a Jew, Jewish thought, Jews as a nation, Israel, the Holocaust, antisemitism, missionaries, Reform Judaism, children and references — plus a separate companion document on Jewish resources by mail order and computer.

Its front matter is more interesting than its index. The document describes itself as an attempt to answer questions that are continually asked on the group, written “by cooperating laypeople from the various Judaic movements”, and instructs the reader in its opening paragraph to make no assumption as to the accuracy or authoritativeness of the answers and to consult a competent authority. It explains that the contributors never standardised on a transliteration scheme from Hebrew, Aramaic, Yiddish or Ladino, so the same word appears with several spellings, and attributes this to the fact that the FAQ “was produced by a committee and is a cooperative work”. It notes where North American movement names differ from those used elsewhere. It states its own goal as presenting a balanced view and flags responses that apply to one movement only.

The dates in its own header are the sharpest thing about it. The heading of the copy posted in March 2004 reads, in the revision-control notation the maintainer left in place, that the last change to the text was made on 19 October 1995 and that the file stood at revision 1.3. Eight and a half years of monthly postings separate the last edit from that copy. That is the shape of the thing: a committee product, versioned and dated, disclaiming its own authority, split into a dozen articles — the guidelines warned that intermediary news and gateway software could choke on large articles, with sixty-four kilobytes “a common magic size” — and posted every month by a script whether or not anything had changed. The group it served has its own page in this directory at soc.culture.jewish.

How much of this literature was soc.*’s is a question the surviving indexes answer less precisely than one would like. Contemporary web front ends to the rtfm.mit.edu archive listed FAQs by newsgroup, and their soc.* branches ran to well over a hundred entries, culture groups predominating with substantial showings from the genealogy, history, religion and support branches. This page has not been able to reproduce a defensible figure from a surviving capture, so none is given here; what can be said from the current ISC files is the size of the constituency, which is 213 soc.* groups, 139 of them under soc.culture, 19 under soc.genealogy, 10 each under soc.history and soc.support. Not every document filed under those groups was crossposted to soc.answers either; the archive also accepted periodic postings that never entered the answers groups at all.

How much passed through it

There is no readership figure for soc.answers and no article count in any document about it. There is, however, a way to get an order of magnitude out of the surviving postings themselves, and it costs nothing but arithmetic. Usenet servers number the articles in each group in sequence and stamp the numbers into an Xref header. Two of the documents quoted on this page were archived with their Xref lines intact, and both were numbered by the same machine, MIT’s own senator-bedfellow.mit.edu.

• The soc.culture.jewish FAQ part one, posted 28 March 2004, is soc.answers:18961.
• The *.answers submission guidelines, posted 4 May 2004, are soc.answers:19061. The introductory posting, sent out in the same minute, is 19063.

One hundred articles in thirty-seven days is about 2.7 a day, or roughly eighty a month, arriving in soc.answers on that server in the spring of 2004. The same two headers give news.answers as 268834 and 270830 over the identical interval: just under two thousand articles, some fifty-four a day. soc.answers was carrying about one article in twenty of what its parent carried, which is what one would expect of a hierarchy filter under a group that received everything.

The absolute numbers invite a second calculation and deserve more caution. If that server numbered soc.answers from the group’s creation in January 1993, then nineteen thousand articles over eleven and a quarter years averages about 140 a month, well above the rate measured in 2004 — consistent with a group past its peak, though a server that joined late, or renumbered at any point, would produce the same appearance. The same 4 May 2004 header ranks the whole family on one machine on one day, and that ranking is worth having:

• news.answers 270,830
• rec.answers 86,889
• alt.answers 72,751
• comp.answers 57,045
• soc.answers 19,061
• misc.answers 17,269
• sci.answers 15,988
• talk.answers 7,078
• humanities.answers 760

Read as a rough census of the reference literature, that puts soc.* fifth of the eight hierarchy groups, behind recreation, alt and computing and ahead of miscellaneous, science and talk, with humanities two orders of magnitude behind everyone. These are one server’s counters, not a measurement of Usenet, and they include whatever that server accepted and later expired; a different site would give different figures. They are offered as the only quantitative evidence about this group that this page could actually verify, and they should be read with all of that in mind.

Eighty-nine control messages

The control-message archive for soc.answers holds eighty-nine messages. One of them created the group. One tried to destroy it. The other eighty-seven belong to a period when the Usenet namespace was under sustained attack, and they make this small administrative group an unusually clear specimen of what that attack looked like from the inside.

Between 26 August and 19 November 1998, sixty-eight forged newgroup messages for soc.answers crossed the network. All purported to come from David C. Lawrence, at isc.org or at uunet.uu.net. All declared the group unmoderated, dropping the moderation marker from the newsgroups-file line they supplied and from the subject line. All cited votes that had never happened, and not consistently: one claims the group “passed its vote for creation by 339:11 as reported in news.announce.newgroups on 27 Aug 1998”, another 139:33 the following day, and the sixty-eight between them invent sixty-eight different tallies, from 322:0 to 50:32. Fifty-two of them carry an X-PGP-Sig header, and all but one of those signature blocks contain underscore characters, which base64 does not use. Five helpfully include a suggested control.ctl configuration instructing administrators to honour every newgroup in a list of hierarchies that includes the non-existent meow.* and to drop every rmgroup unlogged. Every one of them is signed off with a single word: Meow!

That signature places the campaign. The Meow Wars, which began in 1996 and ended around 1998, were a crossposting and flooding campaign whose participants took their name from the word they flooded newsgroups with; among the incidents attributed to them were floods of forged control messages that caused hundreds of oddly named newsgroups to appear on servers in many places at once. soc.answers was not a target in any meaningful sense. It was simply one entry in a list, and the interesting thing about the forgeries is what they attempted: not to delete the group but to redefine it, stripping the moderation flag so that anything could be posted into a group whose entire value lay in the fact that nothing could.

Interleaved with the forgeries, between 7 October and 29 November 1998, are nineteen genuine control messages. They are identical to one another in substance, they restore the moderated flag and the true tally of 240:29 from January 1993, they reproduce the original charter in full, and they carry PGP 2.6.2 signatures over the Subject, Control, Message-ID, Date, From and Sender headers. Somebody was patiently reissuing the truth every few days for two months, in the reasonable expectation that a proportion of the world’s news administrators would apply whichever message arrived last.

The last control message in the file is dated 4 November 2001 and is the only rmgroup ever issued for the group. It is a forgery too, posted over the name of Urs Janßen — the maintainer of the tin newsreader since 1996 — from a host on a Quebec network, with an organisation line borrowed from a students’ association in Karlsruhe. Its entire message body reads: please remove the bogus newsgroup soc.answers. Beneath it sits a signature line of machine-generated nonsense of the kind that period’s forgery tools appended by default. A matching message, differing only in the group name and in its equally random signature line, was issued against rec.aviation.answers twenty-six minutes later.

None of it worked, and the reason is in the configuration file the same archive distributes. The default control.ctl shipped for the Big Eight instructs a server to drop every newgroup and rmgroup for comp.*, humanities.*, misc.*, news.*, rec.*, sci.*, soc.* and talk.* whatever its apparent sender, and to act only on messages from [email protected] that verify against the news.announce.newgroups key, whose fingerprint is printed in the file. Wikipedia’s account of the Meow Wars notes that the control floods prompted Usenet users to adopt digital signatures; whatever the exact chain of cause, PGP is the reason a group created in 1993 by an unsigned message from one man’s workstation was still recognisably itself in 2001.

Why the FAQ existed at all

The problem the FAQ solved is older than Usenet and is not really a technical problem. A public forum with a continuous intake of newcomers receives the same questions forever, and the people best placed to answer them are the people most tired of doing so. Every long-lived community rediscovers this; Usenet’s contribution was to solve it with a document and a schedule rather than with patience.

The acronym itself came from the ARPANET rather than from Usenet. On the SPACE mailing list it had been assumed that new users would fetch and read the archived messages before joining in; in practice many did not, and answering the same questions became tedious enough that participants began distributing standing answers instead. Eugene Miya of NASA coined the abbreviation for that list at some point between 1982 and 1985, and the practice spread to Usenet, where Jef Poskanzer is credited with the first weekly FAQ, in the graphics newsgroups. What began as a private convenience on one mailing list became, within a decade, a genre with a submission process, a header standard, an archive, a mirror network and a moderation team.

The call for votes that created soc.answers records the reasoning of the people who built the repository, quoting the news.answers charter at length, and it is thoroughly practical. Collecting the documents in one group let administrators set a long expiry time on that group alone instead of on every group that happened to contain a FAQ. It let archiving software watch one newsgroup instead of scanning the whole news spool. It gave readers a quick reference they could browse, and a way to find a group’s document by searching headers rather than guessing at subject lines. And it simplified retrieval software. Not one of those four reasons is about the reader’s soul; they are about disk quotas, cron jobs and grep. The hierarchy-level groups were added on the same principle at one further degree of granularity: a site could now take the soc.* documents without the comp.* ones, and a reader could subscribe to the reference literature of the hierarchy they actually cared about.

The same file contains the one flight of rhetoric in the whole cycle, and it is a fair statement of why the archive outlived its source. FAQs, the news.answers charter observes, “are often considered the ‘distilled wisdom’ of a newsgroup or group of newsgroups”, so that a single document from a hierarchy might be of wider interest than the hierarchy as a whole. That turned out to be true in a way nobody intended: the discussions evaporated and the distillate kept.

The evidence for its durability is the mirror list. The introductory posting the moderators sent out monthly carried a directory of archives holding some or all of the collection, sorted by country, and even the copy still going out in 2004 named FTP servers, Gopher servers, mail servers and web sites in Austria, Belgium, Canada, Finland, France, Germany, Japan, Korea, the Netherlands, the United Kingdom and the United States, among them university machines at Vienna, Leuven, Lyon, Paris, Paderborn and Munich. Readers were told it was usually faster and less wasteful of network resources to use an archive in their own country, or at least on their own continent. A literature written by volunteers for individual newsgroups had, without anyone planning it, acquired a distribution network of the kind normally reserved for software.

How it stopped

No announcement marks the end. There was never a posting saying that soc.answers had ceased; there is no rmgroup in the archive except the forged one; the group is still in the newsgroups file and still flagged moderated. What the record offers instead are dates that stop.

The submission guidelines carry a revision control identifier showing their last change on 14 June 2003, and the copy that survives in a web archive was posted on 4 May 2004. The post-approval guidelines were last revised on 9 May 2001 — and the copy of those that survives was posted on 20 May 2006, which is the latest dated *.answers document this page has been able to put its hands on. The apparatus was therefore still running, unaltered, thirteen years after the group was created. This directory’s own gateway, news2mail.com, was carrying soc.* traffic to e-mail during the same era, which is why the group has an address here at all.

The one place where the decline of an answers group is documented in detail is the removal notice for rec.aviation.answers, issued by the Big-8 Management Board in March 2007 after a last call for comments produced consensus. The proponent’s statement is the closest thing the family has to an obituary. Over most of its history the group had carried eight documents. Most were revised at least once a year for the first few years, with changes becoming rarer later, and most had not been updated for roughly a decade. There had been interruptions from March to November 1995 and from May 1997 to February 1998; the final interruption began in February 2002 and had not ended. One document limped on into January 2003. Read against a group created in September 1992, that is a fifteen-year life in which the last five were silence and the five before them were repetition.

What replaced the form is visible inside the guidelines themselves, which is the detail worth keeping. By the version quoted here, the auxiliary header includes a URL: line for the maintainer’s web copy; the moderators note that all postings in news.answers are automatically converted to HTML by, and made available at, several web sites; and the section on posting frequency offers, as a legitimate option, keeping the full document on a web page and only ever posting a brief pointer to it. The system had written its own exit route into its own rulebook. Once a document had a permanent address, the elaborate apparatus of Expires and Supersedes and monthly reposting was solving a problem that no longer existed — the whole point of it had been to keep a copy in front of readers on a medium that threw articles away.

Search engines finished the job. The FAQ’s core promise was that the answer would be somewhere findable; once anything on the web was findable from anywhere, a monthly article in a moderated newsgroup was a slow and lossy way to publish a reference document, and the collaboratively edited encyclopaedia founded in 2001 was in effect the same idea with a better revision model. The form itself has since fallen out of editorial favour on its own merits: the United Kingdom’s Government Digital Service published a note on 25 July 2013 explaining why it does not use FAQs, arguing that the format is convenient for writers and more work for readers. Whether or not one agrees, it is a fair epitaph for a genre whose finest examples — the twelve-part committee documents with their indexes and their disclaimers — had long since stopped being lists of questions at all.

The archive machinery has decayed unevenly. The host name rtfm.mit.edu, which appears in every document quoted on this page, still resolves in the Domain Name System to an address in MIT’s range; no web service answered on it, over either protocol, when this page was written. The documents themselves are in better shape than the plumbing, which is roughly what the *.answers moderators would have predicted: the whole design was aimed at making the text outlive the transport. Where the postings went and how they were filed is covered on the comp.* hub, which owns the story of the archive itself.

The tidiest sentence in the entire corpus is in the post-approval guidelines, under the heading Terminate a posting (i.e., stop posting it forever). It reads, in full: “Inform [email protected].” No procedure, no vote, no five-day objection window. A genre that had built a submission robot, a header standard and an international mirror network had exactly one line of policy for its own ending, and as far as the surviving record shows, most maintainers did not use it.

What the record does not show

Honesty about a group like this means listing the questions the sources do not answer.

• No moderator of its own. The creating control message says the addresses were the same as for news.answers, and nothing in the control archive or the group’s paperwork ever assigns soc.answers a separate moderator. Any name attached to this group specifically would be an invention.
• No readership figures, and no article count from any source that set out to make one. The volume estimates in this page are derived from Xref numbering on a single server on two days in 2004; they are not measurements of Usenet, and no source consulted here reports how many soc.* documents were approved. The List of Periodic Informational Postings ran to twenty parts across all hierarchies; this page has not reconstructed the soc.* portion of it.
• No last article. Nothing establishes when the final document was posted to soc.answers. The dates given above are the last dates on the surrounding documentation, not on the group.
• No file in the proposals archive. The absence of a soc.answers file in the news.announce.newgroups archive is a filing convention, not a gap in the record; the material is under news.answers-expansion. Anyone citing the absence as evidence that the group was undocumented has misread the archive.
• No explanation for the hierarchies that abstained. Why French, Italian, Spanish, Dutch, Australian, Japanese and Scandinavian Usenet never adopted the answers convention is not recorded in any source this page could reach.
• No verifiable count of soc.* FAQs. Web front ends to the archive indexed FAQs by newsgroup in the 1990s, and their soc.* sections were large, but this page could not retrieve a capture good enough to count, so it does not give a number. Indexes of that kind lag, over-count renamed groups and miss documents that were never crossposted.
• Only one FAQ examined at first hand. The description of what a soc.* periodic posting looked like rests on one document read in full with its headers. It is a representative specimen, not a sample.

Scope and limits

This page is built from primary sources that are still online and can be checked line by line: the Internet Systems Consortium’s current newsgroups and active files, its control-message archives for soc.answers and its siblings, its news.announce.newgroups archive, its distributed control.ctl, the *.answers submission and post-approval guidelines and the introductory posting as archived from the web, RFC 1036 and its successors RFC 5536 and RFC 5537, and one archived copy of a soc.* FAQ read in full with its headers intact. Where a figure appears here — thirty-one answers groups, twenty-six of them moderated, 213 soc.* groups, eighty-nine control messages, sixty-eight forgeries, nineteen genuine reissues, twelve parts, 240 to 29 — it was counted from those files rather than repeated from anywhere.

What this page deliberately does not do is retell its neighbours. The invention of the periodic posting, the group-creation ritual of request, call, vote and result, the control message as an instrument, and the fate of the MIT archive belong to the news.* hub. The Great Renaming, the split between soc and talk, and what moderation ordinarily meant in this hierarchy belong to the soc.* hub. The culture branch and its literature have their own hub. This page has tried to own only what nobody else here does: one group’s paperwork, the arithmetic of the family it belonged to, and the anatomy of the documents it carried.

A last caution about register. Several of the documents described above were written by and for communities in the middle of wars, dispossessions and arguments about who counts as a member of what. The FAQs recorded those arguments; some of them were themselves fought over line by line, which is why so many open by disclaiming their own authority. This page describes the form those documents took and the machinery that carried them. It takes no position on anything they said.