Summary
- The hedge survived. The archie manual dated 20 May 1991, a 1992 tutorial, an undated Usenet description and a modern encyclopedic entry all repeat "a list of about 600 Internet ftp archive sites."
- The sentence beside it stopped at the first relay. "there are still some 50 (out of over 600) sites to be entered" appears in both archived copies of the November 1990 announcement and in none of the later texts examined. The first text to drop it is the manual that ends with an attribution to Emtage and Bill Heelan.
One sentence, two modifiers
The announcement distributed on 14 November 1990 does what a carefully reread press release would not. In the body, the software "maintains a list of about 600 Internet ftp archive sites" archived copy at krsaborio. A few lines later, in a parenthesis about the state of the build: "Please note that at the time of writing the database has not been fully built: there are still some 50 (out of over 600) sites to be entered" same message.
The number arrives twice, wearing two hats. As a denominator it is a floor: more than 600, of which about fifty are missing. As a description of the list being maintained, it is an approximation: about 600. Neither occurrence carries a measurement date. The only date anywhere near it is the date of sending.
That this is the text and not a website's layout is settled by a second copy, reposted in comp.archives, which reproduces both sentences word for word and carries an Original-posting-by line in Emtage's name reposted copy. That copy also shows an internal seam worth a sentence: its own date field reads 15 November while the archive name it carries reads 14-Nov-90. One file contradicts itself by a day. That is what archive copies are, and every 1990 quotation in this article rests on one of them. The original server at McGill was not read.
What the first relay dropped
Six months later, the archie manual page dated 20 May 1991 says the same thing in one clean line: "The archie database subsystem maintains a list of about 600 Internet ftp(1) archive sites" archie manual, 20 May 1991.
The hedge is intact. What is missing is everything that made the number honest about itself. The words "over 600" do not appear. The sentence about fifty sites still to be entered does not appear. And this manual is not a third party's summary: its closing attribution names Alan Emtage and Bill Heelan, McGill University same manual.
After that the sentence simply falls out of the record I read. A tutorial text under a 1992 copyright notice reproduces the formula almost identically: "the first maintains a list of about 600 Internet ftp archive sites" 1992 tutorial. An undated Usenet description keeps the number and moves the unit underneath it: "Currently the listings of about 600 sites are stored in the archie database" undated description. That is not the same population as sites on a maintained list, and nothing in the message announces a change. A modern encyclopedic entry, which cites the 1990 announcement by its date, compresses the whole thing into a phrase about "a database of about 600 sites," refreshed nightly at roughly a thirtieth per night tertiary entry.
Four later texts. Four approximations. Zero restorations of the fifty.
210, and the hand left behind
The earlier figure behaves in the opposite way, which is why it belongs in the same article.
The September 1990 message contains the sentence "For what it's worth, we currently know of about 210 sites" September 1990 message. Read the sentence before it, because that is where the hedges live: the team walked the entire list in about a month, so "no entry can be more than that old." Two bounds on one figure — approximate, and no more than a month stale — in the same breath as the number.
That message went out under Peter Deutsch's name, and it speaks of Emtage's scripts in the third person: that is the textual basis for treating Deutsch as the hand that wrote it. The copy carries its own small contradiction too: a header date of 11 September and an archive name of archie/10-Sep-90.
Now watch the two halves separate. Four texts examined connect the September message to Deutsch without printing a single digit. An encyclopedic reference list cites "Deutsch, Peter (11 September 1990)" for a release date and never mentions 210 encyclopedic entry. A chronology dossier asserts that the 10 September date rests on "a Usenet post by Deutsch himself" and gives no figure dossier. Deutsch's own 2000 column claims the posting in the first person and changes the scale: "By 1989 there were dozens of archive sites available on the Internet" Deutsch, 2000. Emtage's 2015 retrospective dates the announcement and speaks of an initial base of "hundreds of repositories," with Deutsch the colleague who asks and then posts Emtage, 2015.
No 210 anywhere in that.
The only text in this corpus that prints 210 is ours. A BTW article published 12 July 2026 states that the 1990 announcement "covered about 210 anonymous FTP archive sites," and attributes it to a document titled "archie: An Electronic Directory Service for the Internet" BTW article, 12 July 2026. That title is not the archived object of the September message, and the article names no author for the announcement.
The correction is therefore ours to make, and it is simple. The figure survived. The hand that wrote it did not travel with it. Two words changed en route as well: the message said the team knew of about 210 sites; our sentence says the service covered about 210, which is a stronger claim about coverage than the 1990 sentence makes.
No date was ever restored
No text examined here attaches a measurement date to the 600. The manual's title supplies 20 May 1991 and says nothing about a recount that month. The 1992 tutorial's only date is a copyright year. The three 1990 sentences in play carry sending dates and nothing else.
That matters, because the two numbers get used as if they formed a series. Nowhere in what I read are they established as the same population: neither message defines what a site is, or says whether the September and November lists count the same thing. My arithmetic here is deliberately trivial — I read 11 September and 14 November off two header lines and note the gap. I do not divide 600 by 210, I do not annualise anything, I do not derive sites per month. Two printed numbers and two dated messages, that is what the record supports.
Why adjacency is the whole story
What is interesting is not that a caveat got lost. Caveats get lost. It is which caveat survived, and why.
"About 600" lives inside the phrase that carries the number. It is grammatically load-bearing: delete it and the sentence must be rewritten. "There are still some 50 sites to be entered" lives beside the number, in a separate sentence, addressed to a reader being told about the state of a build. Copy the description — which is what the manual and everything after it did — and the hedge comes along free, while the confession stays behind.
That is my judgement after reading these twelve documents, not a claim about anyone's intentions. Nobody hid anything. The sentence was there from the start, published, never withdrawn, and easy to find: it sits in the same message as the number it qualifies. It simply was not inside the number's sentence, and sentences are what gets copied.
The practical consequence is uncomfortable for anyone publishing infrastructure counts. A hedge survives if it is inside the unit you cite. A confession of incompleteness survives only if it is welded to the number. Emtage wrote both in one paragraph in 1990, and the copying machinery took one.
Where this record stops
Every 1990 quotation above rests on an archive copy or a repost, never on the original server at McGill. The September mirror prints its own header, the November copies print theirs; a header proves what a copy displayed. The absence findings hold for the twelve documents read on 21 November 2026, not for the whole historical record: there may be later texts that restored the fifty, and if they exist they are not in this corpus. The undated Usenet description could not be dated at all from its header.
The 1992 USENIX table often cited for 1,025 known sites and 886 indexed as of 30 October 1991 was not consulted for this article and is not used as evidence here.
Member Briefing
Read the full trend briefing
Members read the full briefing with the source notes behind every finding.
Strategic Circle
A daily briefing in your language, built from changes we have verified.
Join for USD 20/monthCancel anytime.
Members also get the full research library, a daily briefing on the networks they follow and the private forum.






