Summary
- The first public notice of archie went out on 11 September 1990 under Peter Deutsch's hand, describing Alan Emtage's scripts in the third person and putting the team's knowledge at "about 210 sites" on a crawl that cycled the whole list in about a month.
- Emtage's own 14 November 1990 post printed "a list of about 600 Internet ftp archive sites" one sentence after warning that the database was not fully built and "some 50 (out of over 600) sites" were still to be entered. No reviewed 1990 text states what population or which unit either figure counts.
The launch notice went out in another hand
On 11 September 1990 a message described "an Internet archive server server" to comp.archives. The archive mirror retains a header dated Tue, 11 Sep 90 17:45:06 GMT, while the archive's own name for the file is archie/10-Sep-90. Two identifiers, one document; I print both and reconcile neither (September 1990 announcement, as archived). The author on that copy is Peter Deutsch, and the sentence that ties the software to a person is in the third person: "Alan has a set of shell scripts that automagically calls up some subset of our list of servers each night (to keep load down on any one machine) and do a remote 'ls' on that machine."
Deutsch then printed a caveat with a clock inside it, and after that the figure: "We cycle through the entire list in about a month, so no entry can be more than that old." Then: "For what it's worth, we currently know of about 210 sites." Read together, that is a coverage claim carrying three of the four things a reader needs — an approximate marker, a stated population (sites the team knows of), and a staleness bound of one month. What it lacks is any unit definition finer than "sites" and any measurement date distinct from the posting date.
The same message says what the tool was actually doing: "Currently, archie is pretty brain-damaged. The files are stored as flat ASCII files and the 'prog' command just launches a 'grep' with the appropriate arguments," and "the current system greps through about 31 Meg of listings!!" The way in was "rlogin quiche.cs.mcgill.ca -l listd". Flat ASCII listings searched by grep is a different object from the database announced in November, and I do not treat the two as one thing.
The number, and the sentence that precedes it
On 14 November 1990 Emtage posted "Announcing 'Archie 1.0': The Archive Server Server" (Archie 1.0 post, as archived). A moderated repost in a second archive retains the same wording, and its header names him as the original poster; that header is the authorship evidence held here, and the original message host was not read (repost witness).
The order inside the message is the story. First: "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." Then the number: "Archie is a pair of software tools. the first maintains a list of about 600 Internet ftp archive sites." Then the mechanism: "Each night software executes an anonymous ftp to a subset of these sites and fetches a recursive directory listing of each. We hit about 1/30th of the list each time, so each site gets hit about once a month, hopefully balancing timely updates against unnecessary network load."
So the 600 belongs to a list the tool "maintains", and the same message says a slice of that list was still to be entered. Both sentences are true of the same document and neither is withdrawn. What the document never says is which population 600 counts — sites on the maintained list, sites whose listings had been fetched, or sites actually entered on the day a reader connected.
It also prints a size and a machine: "Even though the database is now about half the size of the version of archie that we have been running for the past couple of months", and "it still occupies about 40 Mb and the updates and searches can still put a noticable load on a Sun 4/280 which is doing little else." The misspelling is reproduced as printed. The hardware travels with the figure; the unit does not.
Three sizes, no unit
September's number measures "about 31 Meg of listings" in flat ASCII files searched by grep. November's measures "about 40 Mb" for a database on a Sun 4/280 that was "doing little else". A third copy of the same November announcement prints "about 55 Mb" where the other copies print 40 (variant copy). Nothing retained here establishes that copy's date or its authorship line, so whether it is a revision, a second copy of one message, or a transcription difference is unknown — and I print the split rather than calling it a correction.
No reviewed text states what any of the three measures: stored listings, searchable database, compressed ftp copy, or the index as a whole. That is why I refuse the ratio. Thirty-one Meg is not "the old database" and 40 Mb is not "the new one" in any sense that licenses arithmetic between them. The same message also compares the database to "about half the size of the version of archie that we have been running for the past couple of months" — a comparison to a baseline it never prints.
Where my arithmetic stops
One subtraction is available and it is mine. "Over 600" less "some 50" gives about 550. That is arithmetic on two printed numbers in one paragraph — it is not a count of indexed sites, not a count of entered sites, and not a claim about what any user could search in November 1990. The distance between the two documents' own header dates, 11 September and 14 November, is about nine weeks. It is not a growth rate: no reviewed text says the two figures count the same population, and a rate calculated across unstated units is a number wearing a unit it did not earn.
A third formulation, and a three-way date
Emtage's own retrospective dates the launch himself: "On September 11, 1990 we announced Archie, the world’s first Internet Search Engine." (retrospective). On the account retained for that page, the initial database covered files on "hundreds of repositories" — a third formulation that matches neither 210 nor 600, and "hundreds" is not a count. I reproduce it as printed and will not promote it to the authoritative figure.
A third-party encyclopaedia entry puts the release at 10 September 1990 and describes the earliest archie as searching "a list of public anonymous File Transfer Protocol (FTP) sites using the Telnet protocol" (entry). That is one day before the author's own date and one day before the archived header of the launch post, and the archive's file name says 10-Sep-90 as well. The dating of the launch is a three-way seam in the record; no figure in this piece rests on the encyclopaedia.
What an index owes beside its number
A public index publishing a coverage figure owes four things beside it: whether the figure is approximate or exact, the denominator it belongs to, the unit of the count, and the date of measurement as distinct from the date of posting. Emtage's November message carried the first and, unusually, the second — fifty of over six hundred, in its own sentence. It carried neither the unit nor a measurement date.
That is the sharp end of this record, and it is narrower than a scandal. The caveat was printed. It was printed as its own sentence, ahead of the number, where any quotation of the number alone can leave it behind without altering a word of what remains. Nothing in the reviewed documents claims the figure was inflated, and nothing here argues it. The finding is about where the qualifier sits, not about why it went missing.
The number survived the paragraph it came from. Anyone quoting it now can put the fifty back — and should say, in the same breath, that no 1990 document tells us what the six hundred counted.
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.






