microsoft.public.fr.scripting.ie-oe-animation

Long-form reference · 7,969 words · about a 35-minute read

The one line of paperwork this group has

Most of the groups in this directory arrive with at least one administrative document behind them: a control message, a charter, a vote result, something written down by somebody at the time. This one arrives with less than that. The Internet Systems Consortium distributes two files to news administrators — an active file, which lists the names a server may carry and whether posting is permitted, and a newsgroups file, which pairs each name with a one-line description. Both still carry this group. Against microsoft.public.fr.scripting.ie-oe-animation the description file prints four words: French Microsoft scripting newsgroup. The active file carries the same name with the flag y, meaning unmoderated and open to posting, and with the article watermarks 0000000000 and 0000000001 — placeholder values that the file prints against every one of its entries, and which say nothing whatever about how much was ever posted.

What is absent is the control message. The consortium also mirrors Usenet's control traffic group by group, and its archive for the microsoft hierarchy is substantial: files for 3,991 distinct names, 3,118 of them beginning microsoft.public. Neither this group nor its parent microsoft.public.fr.scripting is among them. There is no archived newgroup message for either name, no archived removal, and no archived reissue. The name reached the master description list that news administrators around the world load into their servers, and reached this directory's catalogue through the gateway that produced it, and the step in between is not in the record.

That gap is worth stating without drama, because it is common rather than singular. Of the 1,770 microsoft names the description file still lists, 382 have no control file in the archive at all; in the French branch the proportion is higher, at 30 of the 103 names. Two-thirds of that French cohort name products that arrived in the 2000s — the eight-group Vista family, the six-group Windows Server family, Internet Explorer 7, Tablet PC, Windows Mobile, Windows Live Messenger — which suggests a control channel that fell behind a catalogue still being extended, rather than anything peculiar to this group. It also means the absence of a control message here proves nothing at all about whether the group was properly created.

Nor is the description this group's own. The identical four words stand against the parent microsoft.public.fr.scripting, because the descriptions in that file were generated by family rather than written by group. The same mechanism produced some genuinely comic results elsewhere in the French branch: microsoft.public.fr.accessoires is described as a French Microsoft Office Access newsgroup, on the strength of the first six letters of its name, and microsoft.public.fr.securite_sociale — created on Microsoft's news servers in 2007 for the French social-security sector — is filed in the catalogue as a French Microsoft health insurance newsgroup, a line that can be read verbatim in the control article that propagated it. A description in that file tells a reader which family a name was filed under. It does not tell them what the group was for, and for this group it does not even distinguish it from its parent.

Everything else on this page is therefore context rather than testimony: the shape of the branch the group belonged to, the technology it was named after, and the two applications whose names are welded together in the middle of it. That is a large body of verifiable material, and none of it is a substitute for the traffic, which has not been examined here.

One hundred and three French groups

The count is checkable and it is worth stating precisely, because nobody appears to have published it. In the master newsgroups file as the Internet Systems Consortium distributes it today, the microsoft hierarchy has 1,770 names. Of those, 103 begin microsoft.public.fr. — and this group is one of them.

Set that number against the rest of the namespace and something about the hierarchy's design becomes visible at once. Sort every third-level component in microsoft.public.* by how many names hang beneath it, and the top of the list is not a product. It is a language: de with 129 names, fr with 103, es with 93. The first product branch, windows with 72, comes fourth; the next, dotnet with 49, comes seventh, behind jp with 53 and it with 50. Below them the pattern holds: cn 49, nl 43, tr 40, br 36, tw 26, pt 25, il 25, arabic 25, kr 24, pl 17, ru 15. Eight of the ten largest branches of a corporate hierarchy of 1,770 groups were national ones. How the namespace was constructed and governed, and what became of it, is set out on the microsoft.public.* hierarchy page and is not retold here.

What the 103 French names cover is a fair miniature of the whole product line, rebuilt in a language. The office applications are all there — access, excel, word, powerpoint, publisher, project, visio, infopath, frontpage, sharepoint, officexp, officedev. So are the servers and the business systems: exchange with five sub-groups of its own, sqlserver, iis, isaserver, sbs, sms, mom, navision, crm, biztalk, update_services. The developer side runs from vb, vc, fox, ssafe and vstudio through a full dotnet sub-branch with rooms for ADO.NET, ASP.NET, C# and Visual Basic. The operating systems are stacked by version in the way the whole hierarchy stacked them: windows98, windowsme, windows2000, windows2000server, windowsce, a six-group windowsxp family with separate rooms for connection, beginners, installation, hardware and multimedia, and later an eight-group Vista family.

Around that core sits the miscellany that makes a national branch look national. There are games — ageofempire, flightsim, trainsim, jeuxmicrosoft, xbox; consumer software — money, works, mappoint; an accessibility group, handicap.accessibilite; a group for Microsoft products on the Macintosh; two for MSN services; and the public-sector groups discussed below. And there is this one. Scripting is a family of exactly two names in French — this group and the parent it cannot be distinguished from by its description — where the browser and the mail client each ended up with rooms of their own.

Counted in full, 63 of the 103 names have four components, 26 have five and 14 have six — that is, one, two or three components hanging below the fr branch itself. This group, at six, is among the deepest names France was given.

The administrative record of the French branch

The group itself left no control messages, but the branch around it left a great many, and they are the nearest thing to a documented history the French side of the hierarchy has. The consortium's control archive holds files for 163 names beginning microsoft.public.fr. — sixty more names than the description file still lists, which is the ordinary arithmetic of a namespace that was pruned more often than it was catalogued. Those files contain 310 control articles between them, and reading all of them takes an afternoon and settles a surprising number of questions.

The earliest are dated 9 April 1998, and there are nine of them, sent between 11:35 and 14:07 GMT: access, excel, money, vb, sqlserver, ie4, windowsnt, windows95 and outlook. They did not come from Microsoft. Every one of them was injected from a news host belonging to a Swiss company — the messages give Cerberus AG as their organisation and mthree.cerberus.ch as their posting host — carried the header Distribution: collabra-internal, and consisted of the two lines such a message is meant to carry: the words For your newsgroups file, and the bare name with no description after it. Two other outsiders follow in 1999: the news administrator of a Swiss internet provider creates the French Windows 98 group in March, and an address at a company called Concentrex creates the Word and Small Business Server groups in October, again with the Collabra distribution header. That is how a vendor hierarchy actually propagated in the late 1990s — not by decree from the vendor, but by whichever administrator on the network took the trouble to ask everyone else to create the names.

Microsoft's own address does appear, and dominates the record: [email protected], an operations mailbox on a Microsoft news host, accounts for 226 of the 310 control messages in the French files. The largest single day is 27 November 2000, when newgroup messages went out for 63 French names inside two minutes, at 10:39 and 10:40 Pacific time — part of a reissue of the whole hierarchy, since the same timestamps appear against English groups too. Reissuing was not a change of policy but of housekeeping: a control message is a request that every administrator answers privately by configuration, and sending it again improves a name's odds of being carried.

The most revealing single day is 2 November 2001, when 34 French names were removed at once by a party with no connection to Microsoft at all. The messages came from Ed Hew of XeniTec Consulting Services in Kitchener, Ontario — the organisation is named in the headers — over the signature usenet hierarchy administration, and each stated its grounds in the body: a control message to counter numerous forged newgroup commands, and remove bogus newsgroup — the name following — which was never created in accordance to the accepted rules and methods for this hierachy, the misspelling being the original's. The notice added that, given the actions of the forger (over 100 forgeries so far), the message would be repeated at reasonable intervals until the problem was dealt with. Thirteen of the 34 names are in the master description list to this day — access, excel, word, outlook, money, iis, sqlserver, vb, vc, frontpage, flightsim, ageofempire and the Exchange setup group — which is the whole lesson of Usenet's control channel in one line: nobody could remove a name, only ask.

The last French newgroup articles in the archive are dated 4 April 2007 — administration_centrales, collectivites_territoriales and sante, for central government, local authorities and health — with securite_sociale following on 6 June 2007. These arrived PGP-signed from a private individual, Julien Élie, at the address [email protected], and they say plainly what they are doing: the group has recently been created on Microsoft news servers so I send this PGP-signed control article to improve its propagation. That division of labour is not an anomaly. INN's control.ctl, the file that tells news software whose control articles to obey, still carries an explanatory note against the Microsoft hierarchy: control articles for it are not issued by Microsoft itself but by a Usenet active participant in order to improve the quality of the propagation of Microsoft newsgroups. It records the administrative group as microsoft.public.news.server and the syncable server as msnews.microsoft.com, instructs servers to drop any unsigned newgroup or rmgroup in the hierarchy, and names the key to verify against. A francophone volunteer signing the control channel of an American corporation's newsgroups is a fair emblem of how the whole arrangement worked. Of those four groups, incidentally, only securite_sociale is still described in the master list — which is why the catalogue's oddest French line survives while the three created alongside it do not.

The last entries of all are removals: 34 of them, sent in a single minute on 15 December 2009, again PGP-signed from the same volunteer address. What happened to the hierarchy after that, and where the readers were sent, belongs to the microsoft.public.* hierarchy page linked above.

One group where the English hierarchy had a family

The name is the most informative thing this group possesses, and it repays being read slowly. microsoft.public is the fixed stem; fr is the language axis; scripting is the subject family; and then, instead of a further component, comes a single leaf that welds three separate things together with hyphens: ie-oe-animation. Internet Explorer, Outlook Express, animation.

Hyphens are rare in this hierarchy. Of its 1,770 names, exactly fourteen contain one, and most of those are compound product or topic names — streets-trips, hyper-v, mac-specific, queries-sql, help-and-support, isapi-dev, msxml-webrelease. This is the only French one, and it is the only name in the hierarchy that contains two hyphens, and so the only one that uses them to join three subjects rather than two. The nearest parallel is Turkish: microsoft.public.tr.oyun-donanim joins games to hardware in the same way, with one hyphen and two topics. Somebody here had three subjects, one group's worth of expected traffic, and no appetite for three group names.

What that compression looks like is clearest against the English side of the same hierarchy. The control archive holds files for microsoft.public.windows.inetexplorer.ie5.programming.dhtml and, beneath it, separate names for .authoring, .behaviors, .scripting and .scriptlets; a matching ie55 family with four names of its own; microsoft.public.inetexplorer.ie5beta.programming.dhtml; microsoft.public.inetsdk.programming.dhtml_editing; and microsoft.public.inetexplorer.ie4.scripting, created on 13 November 1998 and finally removed on 16 June 2005. The general English scripting family that survives in the master list has eight members — debugger, hosting, jscript, remote, scriptlets, vbscript, wsh and virus.discussion. The English-language reader who wanted to ask about a DHTML behaviour had a group for behaviours. The French-language reader had this one.

France did briefly get something more conventional, and its life is one of the shortest in the archive. On 28 August 2000 at 09:28 Pacific time, Microsoft's operations address created microsoft.public.fr.devasp, microsoft.public.fr.devdhtml, microsoft.public.fr.devxml and a general microsoft.public.fr.ie. Three days later, on 31 August at 10:47, all three dev groups were removed again, and over the next four minutes eight new names went out: five that replaced them — devaad, devadsi, devcom, devinternet and devmultimedia — and three others carried along in the same burst, among them, at 10:49, a French Outlook Express group of its own. The French DHTML group had existed, as an administrative fact, for three days. Whatever the reasoning — and none is stated in the messages, which carry no text beyond the catalogue line — the effect was that French web-development questions were consolidated into a general internet-development group at the end of August 2000, that the browser and the mail client each acquired a general room (joined a year later, on 7 September 2001, by ie6 and outlookexpress6), and that the group named on this page was left holding the specifically dynamic-and-animated end of the subject.

The third element of the name is the one that dates the group's subject most precisely. Animation was not a loose description; it was a heading in Microsoft's own documentation. The company's Web Workshop pages for Dynamic HTML, as archived, list the subject's topics as Dynamic Styles, Dynamic Content, Positioning and Animation, Filters and Transitions, Font Download, Data Binding and the DHTML Object Model. A group called ie-oe-animation is a group named out of the vendor's own table of contents.

A newsgroup about scripting the program its readers were most likely reading it in

The middle two letters of the leaf name are the oddest thing about this group, and the reason the page is worth keeping. Outlook Express was not only a mail client. It was a combined mail and news client — the software's own genre, in the reference literature, is given as e-mail client and newsreader — and it arrived on an enormous number of desktops without anybody choosing it.

The bundling history is precise. The program began in 1996 as Microsoft Internet Mail and News, version 1.0 shipping alongside Internet Explorer 3 and version 2.0 later the same year; the executable kept the name msimn.exe to the end, a fossil of that first title. In 1997 it was rebuilt, renamed Outlook Express and bundled with Internet Explorer 4. Outlook Express 4.0 came with Windows 98 in June 1998; 5.0 with Windows 98 Second Edition in June 1999; 5.01 with Windows 2000 in February 2000; 5.5 with Windows Me in June 2000; and 6.0 with Windows XP in October 2001, integrated with Internet Explorer 6. It was bundled with versions of Windows from Windows 98 through to Windows Server 2003, and was available for Windows 3.x, Windows NT 3.51, Windows 95 and the classic Mac OS. It is a different application from Microsoft Outlook, shares no codebase with it, and was never a cut-down version of it, whatever the names suggest.

So the reversal is this, and it is worth stating once and exactly. This is a French-language newsgroup about scripting Internet Explorer and Outlook Express, and a large but unquantified proportion of the people reading it were reading it in Outlook Express — because on a Windows machine of that period the newsreader was already installed and already configured for the mail account. Microsoft ran the server the hierarchy synchronised from, msnews.microsoft.com, which the control file still names; how readers were pointed at it is on the hierarchy page. How many of this group's readers actually used that client is not established anywhere and is not asserted here; the surviving article headers of the microsoft hierarchy would settle it and have not been examined for this page. What can be said without a number is that the application under discussion was, for a great many Windows users of the late 1990s and early 2000s, the program through which Usenet was experienced at all — and that a question in this group about how a script behaved when Outlook Express rendered a message could be answered by a reader watching that behaviour in the window the question had arrived in.

What DHTML actually was

Dynamic HTML was never a specification and never a product. It was a term used by browser vendors for a combination — markup, style sheets and a client-side scripting language, joined by an object model that let the script reach into the document after it had loaded and change it without going back to the server. Microsoft introduced the term with Internet Explorer 4, released on 22 September 1997, and its own documentation was admirably blunt about the scope of the claim: Dynamic HTML (DHTML) is a set of innovative features in Microsoft Internet Explorer 4.0.

The pieces, as that documentation set them out, were these.

• Dynamic styles. Every element's inline style was reachable from script as an object, so a script could change a colour, a font size or a display value and the browser would reformat and repaint immediately. Hiding a list until the reader clicked something needed no second page and no server round trip.
• Dynamic content. Properties and methods for rewriting the document itself — replacing an element's markup, replacing its text, inserting new markup adjacent to an existing element. One of these, innerHTML, outlived everything else in this list: Internet Explorer introduced it, every other browser copied it for compatibility, and the standards caught up with it only in HTML5.
• Positioning and animation. An element could be given an x-coordinate, a y-coordinate and a stacking order, and — because the object model reached the style properties — those coordinates could be changed after the page had loaded. Animation in this world was not a media format. It was a script that adjusted a position slightly, on a timer, until the element arrived where it was meant to be. That is the animation in this group's name, and the entire technique fits in a paragraph.
• Filters and transitions. An extensible architecture for visual effects applied to elements and for transitions between states — proprietary, and never adopted elsewhere.
• Font download, and data binding. Fonts fetched with the page; and a mechanism by which a data source object supplied records that the page's elements were bound to, so that a table could be re-sorted or re-filtered on the client. Data binding was the most ambitious and least portable idea in the set.
• The object model. The foundation of the rest: every element in the document available as an object with properties and methods, all text reachable through those objects, and user actions — a key pressed, a mouse moved or clicked — delivered as events that a handler function could intercept.

Two languages could drive it. Microsoft's documentation named them together in the same sentence: Microsoft JScript, described there as compatible with the ECMA 262 language specification, or Microsoft Visual Basic Scripting Edition. JScript had first shipped in Internet Explorer 3 in August 1996 and was implemented as an Active Scripting engine, which is why the same engine could be plugged into Active Server Pages and the Windows Script Host as well as the browser. VBScript, the Windows Script Host and classic ASP are the subject of microsoft.public.scripting.vbscript and are not covered here. The point for this group is narrower and slightly awkward: a French developer writing DHTML chose between two languages, only one of which any other browser would run.

The other model, and the standards that arrived afterwards

Both major browsers shipped a fourth version in 1997 and both called what they had added dynamic HTML, and the two things were not the same thing. Netscape Navigator 4 had a dynamic system, but not every element in the document was represented in its object model; the unit of dynamism was a layer — a non-standard element, or any absolutely positioned element, each treated as a separate document object, addressed through a document.layers collection and filled by writing into it. Internet Explorer 4 exposed the whole document instead, through an all collection on the document object, so that any element with an identifier could be reached and altered. Microsoft's documentation described its object model, correctly and rather pointedly, as a superset of the JavaScript Object Model found in Netscape Navigator.

The consequence for anyone writing a page was the branching idiom of the period: test whether document.layers existed, and if it did not, assume the other model. The test was used to decide which browser was running even in scripts that had nothing to do with layers at all. The reference works call these two incompatible generations the intermediate DOM — after the legacy model of forms, links and images that both browsers had inherited from the first generation of JavaScript, and before anything standardised existed.

The event models diverged in the same way and more deeply, because the disagreement was about direction. Where two elements are nested and both are listening for a click, something has to decide which hears it first. Netscape's model ran the event from the outside inwards — capturing. Internet Explorer's ran it from the inside outwards — bubbling. Each vendor's handlers were registered by its own methods, attachEvent and detachEvent on one side against the addEventListener and removeEventListener that the standard would later adopt. A script that wanted to work in both had to do everything twice.

The standards were arriving in parallel and mostly behind, and the dates are worth having in order. Cascading Style Sheets level 1 became a Recommendation on 17 December 1996 and level 2 on 12 May 1998. Positioning — the specific thing this group was named after — travelled separately and revealingly: Positioning HTML Elements with Cascading Style Sheets was a W3C Working Draft, edited by Robert Stevahn of Hewlett-Packard and authored jointly by Scott Furman and colleagues at Netscape Communications and Scott Isaacs and colleagues at Microsoft. Its 19 August 1997 draft is the one the browsers of that autumn were implementing; a version dated 31 January 1997 preceded it, a further revision followed on 2 September 1999, and the whole document was ultimately superseded by CSS level 2. Microsoft's own DHTML pages say so explicitly, stating that dynamic positioning and animation are based on that Working Draft and will be adapted, as needed, to conform to the final recommendation. Whatever else the two companies were doing to each other in 1997, their engineers were co-authors of the draft that made positioned, animated elements possible.

The Document Object Model proper came later still. DOM Level 1 became a W3C Recommendation on 1 October 1998 — a year and nine days after Internet Explorer 4 shipped, by which time both vendors had two generations of incompatible object models in the field. It described a document as a tree of nodes that a script could read and modify, and it said nothing at all about events. Events had to wait for DOM Level 2 Events, a Recommendation of 13 November 2000, which settled the old argument by refusing to choose: an event travels down from the root to the target, is delivered at the target, and then, optionally, bubbles back up. Capture, target and bubble — three phases, one from each side of the dispute. DOM Level 3 followed in April 2004 and added keyboard event handling and XPath.

Not everything Microsoft built in this period was rejected and not everything was accepted. The company put the behavior style property — the mechanism behind HTML Components, which let an element's behaviour be defined in a separate file — before the W3C, along with the HTML+TIME timing profile and the Vector Markup Language; all three were published as W3C Notes in 1998 and none was adopted as it stood, though VML was combined with a competing proposal from Adobe and Sun to produce the working group that made SVG. innerHTML and XMLHttpRequest, by contrast, were copied by everyone and standardised eventually. The commercial and legal history of the two companies' rivalry — the market shares, the bundling, the court record — belongs to alt.netscape.sucks and none of it is repeated here.

One last date makes the timing of this group's subject intelligible. ECMAScript's first edition was adopted in June 1997, three months before Internet Explorer 4 shipped; the second edition followed in August 1998 and the third — the version most people mean when they say the language of that era stabilised — in December 1999. The scripting language, the style-positioning draft, the object model and the event model were all in motion at once, in different committees, at different speeds. That is the world the group was created to answer questions about.

Scripting the two applications

The two halves of the group's name were scriptable in quite different senses, and the asymmetry is the most useful technical thing this page can record.

Internet Explorer was scriptable in every direction. Inside a page, the object model described above gave a script the document. Outside it, the browser was built as a set of Component Object Model components in separate libraries: the layout engine, MSHTML, which debuted with Internet Explorer 4, was designed from the start as a software component that other programs could host, presenting a COM interface for reading and editing a page in any COM-capable environment. A WebBrowser control could be dropped into an application, driven from that application, and its events captured. Browser Helper Objects — plug-in libraries introduced with version 4 in October 1997 and loaded into each new instance of the browser — extended it from the other side. And the same Active Scripting engines that ran JScript and VBScript in a page also ran them in the Windows Script Host and in Active Server Pages, so that a technique learned in one place transferred to the others.

Outlook Express was not scriptable in that sense at all, and this is the point at which the group's name promises more than the software delivered. Microsoft never published an object model for it. Except for Simple MAPI messaging, its programming interface was undocumented, and programmatic access to it or control of it was neither documented nor supported; two storage interfaces, IStoreNamespace and IStoreFolder, were documented in 2003, and they concern the message store rather than the application. Add-ins for it existed, but were built by hooking an interface the vendor had not described, using Windows hooks — the best-known example being a PGP plug-in whose source was published under the GPL, which is how the technique became known at all. Microsoft Outlook, the other product with a similar name, did have a documented object model, and the successor to Outlook Express would ship one. Outlook Express, in the years this group existed, did not.

What Outlook Express did have was a rendering engine, and it was Internet Explorer's: MSHTML drew its HTML messages. Everything the browser could do with a document, the mail window could in principle do with a message: styles, positioning, the object model, and script. The scriptable surface of Outlook Express was therefore not an automation API but the message body itself — stationery, signatures, HTML mail composed with dynamic effects in it. That is a strange kind of scriptability: not write a program that drives the mail client, but write a document that the mail client will execute when it draws it. The consequences of that distinction occupy the next section.

The boundary that was supposed to govern all of it was the zone model. Internet Explorer grouped content by origin and applied restrictions per group; the registry that stores the settings defines five zones numbered zero to four — My Computer, Local Intranet, Trusted Sites, Internet and Restricted Sites — each with its own policy for downloading and running ActiveX controls, running scripts, and much else besides. A separate marking governed individual controls: a control could be flagged safe for scripting, meaning a page could instantiate and drive it without asking the user's permission. The Win32 security model underneath all of this is the subject of microsoft.public.platformsdk.security and is not covered here.

The reversal: what mail-and-news scriptability was actually used for

Everything described above — an object model reaching every element, a scripting engine shared between the browser and the shell, controls marked safe to script, and a mail client that rendered messages with the browser's engine — was designed as convenience for authors. Between 1999 and 2001 it was used, repeatedly and at scale, as a delivery mechanism. The record is set down here as history; nothing in this section is advice, and no working technique is described.

The Melissa virus, released on 26 March 1999, was a macro virus in a Word document rather than a browser script, but it established the pattern the rest followed: on opening the document it used Microsoft Outlook to send copies of itself to the first fifty entries in the user's address book, so that the messages arrived from a known correspondent.

Later in 1999 came the one that belongs to this group's subject exactly. The Kak worm was written in JavaScript and spread through a flaw in Outlook Express, catalogued as CVE-1999-0668. It exploited an ActiveX control that had been marked safe for scripting on the reasonable-looking grounds that it created type-library files, and which turned out to place no restriction on what those files contained or what extension they were given. Kak placed its code in the user's e-mail signature, so that it ran when a message was viewed or previewed — because Outlook Express used Internet Explorer to draw the message — and used the control to write a file into the startup folder, where it would be run by the scripting host at the next boot with more privilege than a page had. No attachment had to be opened. This is the reversal of the group's subject in its purest form: a script in a mail message, executed by the newsreader, on a machine whose owner had done nothing but look.

ILOVEYOU followed on 4 May 2000. It was written in VBScript, ran under the Windows Script Host, arrived as an attachment named to look like a text file, overwrote files with copies of itself and mailed itself onward using the addresses it found on the machine. Tens of millions of infections were reported within ten days of the first cases, and it is estimated that a tenth of the internet-connected computers in the world were eventually affected; the damage figure usually quoted in recent accounts is around ten billion US dollars. Philippine prosecutors dropped all charges against its author, a college dropout in Manila, for want of any law against hacking in the country at the time; an electronic-commerce law signed that June could not be applied retroactively.

Then a small piece of evidence from the hierarchy's own paperwork, which is the sort of thing this directory exists to notice. On 5 May 2000 at 10:00 Pacific time — the day after ILOVEYOU began spreading — a control message went out from Microsoft's operations address creating a new group: microsoft.public.scripting.virus.discussion. It sits in the English scripting family to this day, between vbscript and wsh. The company that had built the scripting engine opened a Usenet group about scripting viruses the day after the most destructive demonstration of what one could do.

The pattern continued. The Anna Kournikova worm of 11 February 2001 was written in a few hours with a VBScript worm-generator toolkit by a twenty-year-old in the Netherlands, and mailed itself to every address in the recipient's address book. Nimda, first reported on 18 September 2001, used five infection vectors at once, of which e-mail was only one — the others being open network shares, browsing of compromised web sites, directory-traversal flaws in Internet Information Services, and back doors left by earlier worms. By then the argument about default behaviour had been settled by events rather than by design.

The defaults changed accordingly, and the dates are on the record. Outlook Express Service Pack 2, delivered as part of Windows XP Service Pack 2 in 2004, stopped loading images in messages by default, began warning about potentially malicious attachments, and — the change that matters most here — used only the Restricted Sites zone for HTML mail, the most constrained of the five, in which script does not run. The same service pack added the Attachment Execution Service, which marks files that arrive from the internet or by mail so that Windows checks their origin and warns before one is opened. The scriptable mail client, in other words, was answered by switching the scripting off where mail was concerned, and by making the zone model do the work it had been designed for. Nothing about the object model itself changed; what changed was which zone a message was judged to belong to.

Writing it in French

The specific difficulty of this group's audience was not that the documentation was in English — a good deal of it was translated — but that the machine was bilingual in the wrong places. The object model was English and stayed English: document, style, innerText, onmouseover, srcElement. The error messages were French. The keyboard was French. The variable names, for the first two and a half years of the language's standardised existence, could not be.

That last constraint is precise and can be read straight out of the specification. The first edition of ECMA-262, adopted in June 1997 — three months before Internet Explorer 4 shipped — says in section 7.5 that each character of an identifier must be a letter, a decimal digit, an underscore or a dollar sign, and then gives the grammar exhaustively: the twenty-six lower-case and twenty-six upper-case characters of the unaccented Latin alphabet, plus the dollar and the underscore. Not a single accented character is admitted. The same document, in section 4.3.16, defines a string value as a sequence of Unicode characters. A French developer in 1998 could therefore put é in the text a script displayed but not in the name of the variable that held it. The restriction was lifted in the third edition, published in December 1999, whose section 7.6 interprets identifiers according to the grammar in the Unicode standard and records exactly one departure from it: the dollar sign and the underscore are permitted anywhere in an identifier. The habit outlived the rule by a wide margin, as habits do: code written for a French audience went on being written with unaccented identifiers long after accented ones became legal, because nobody could be sure which engine would meet them.

The keyboard compounded the problem in a way that shows up in code rather than in prose. The French AZERTY layout gives keys of their own to five accented letters — é, è, à, ù and ç — and puts the digits behind the shift key, and it deals with the remaining punctuation by putting it on a third level reached with AltGr, printed at the bottom right of the key. Read the standard layout across the number row at that third level and it is, in order: tilde, hash, brace-open, bracket-open, pipe, grave accent, backslash, caret, at-sign, bracket-close, brace-close. Every one of those characters is structural in JScript and in HTML. A francophone developer opened a block with a two-key combination, closed it with another, and typed a backslash the same way. This is not a hardship; it is a small, permanent tax on writing code, levied on one population and not on another, and it is the reason keystroke instructions written by a helpful English-speaking correspondent so often did not survive the journey. The keyboard as a language question, and French Usenet's long argument about whether accents survived the network at all, belong to alt.fr.comp.os.ms-windows.xp and are not retold here.

The localisation of the applications produced the sharpest practical difficulty, and it is one that any support group in a language other than English will recognise. When a script failed, the browser and the scripting engine produced a message in French. The English-language reference documentation, the vendor's knowledge base articles, the English newsgroups where the same bug had probably already been diagnosed, and eventually the search engines, were all indexed on the English wording. A French error message was simultaneously the most useful thing the developer had and the least searchable. The natural move — translate the message back into English and search for that — required knowing the official English string, which is exactly the knowledge the person asking did not have. A French-language group was not a cultural preference in that situation. It was the only place where quoting the message one actually had in front of one produced an answer.

There is a small irony in the catalogue itself. The line that has kept this group addressable, quoted in the first section above, is written in English: French Microsoft scripting newsgroup. The group was French. Its description, like its object model, was not.

What became of it all

The applications went first, and they went in the order they had arrived. Outlook Express 6 was the last version to carry the name; Windows Vista replaced it with Windows Mail, released with that system on 30 November 2006. Windows Mail was a genuinely different program — a transactional database in place of the single file per folder that had made Outlook Express notorious for corrupting mailboxes, account settings in XML files rather than the registry, and, tellingly, a documented COM programming interface where its predecessor's had been undocumented. It kept the newsreader: messages and newsgroup articles were stored as separate files, .eml and .nws. Windows Live Mail, first released on 6 November 2007 and built by the same team, kept it too — mail, a calendar, an RSS reader and a Usenet newsreader in one program. The Windows Essentials suite containing it reached the end of its support on 10 January 2017 and was withdrawn from download, installed copies continuing to work unsupported. That is as close as the lineage has to an end date: the last Microsoft mail client to include a Usenet newsreader lost its support in January 2017.

Internet Explorer took longer. New feature development ceased after Microsoft Edge arrived; Internet Explorer 11 was the final release; support for it ended on 15 June 2022 for the Semi-Annual Channel editions of Windows 10. The engine itself is not finished: the core of Internet Explorer 11 is committed to shipping and receiving security updates until at least 2029 as the IE Mode of Edge, which is to say that the layout engine this group was named around is contracted to outlast the browser by seven years and to still be running more than thirty years after the version that introduced DHTML shipped.

The technology's name faded before either application did. The reference works record that the term DHTML fell out of use because it had become associated with practices that did not work reliably across browsers — which is a polite way of saying that once the object models converged, there was nothing left for the word to distinguish. What the technique described did not fade at all: changing a loaded document from script is simply what web pages do now, and the successor idea, adding network requests after load, arrived under a different acronym.

Where the questions went is partly documented and partly not. The hierarchy's own closure, the web forums that replaced it and the migration Microsoft published are covered on the microsoft.public.* hierarchy page. The wider migration of programming questions to a general question-and-answer site began with Stack Overflow, created in 2008. In French, one of the standing destinations was the francophone developer community Developpez.com, whose French Wikipedia entry — which carries a warning that the article's claims need sourcing — dates its founding to 1999; if that is right, it was already running for most of the years this group existed, which is a reminder that the newsgroups were never the only room in the building.

What is left of the branch is the catalogue. All 103 French names are still in the file that news administrators load, this one among them, each with its family-level description, none with any traffic behind it. A modern server that syncs its group list from that file will offer a French-language Microsoft scripting group to anyone who scrolls far enough — which is, in a small way, why this page has to keep resolving.

The group's own record

It is worth separating what the record establishes from what a reader might reasonably assume, because for this group the gap is unusually wide.

Established: the name exists, and existed long enough and widely enough to be carried into the master files that the Internet Systems Consortium distributes. It is listed as unmoderated and open to posting. It sits under the French language branch of a corporate hierarchy of 1,770 names, in a family of exactly two — itself and its parent — where the equivalent English family has eight and the equivalent English DHTML family under Internet Explorer 5 had five names of its own. Its leaf name joins three subjects with hyphens, a construction used in only fourteen of the hierarchy's names, in no other French one, and with two hyphens in no other name at all. Its subject can be dated closely: the three things it is named after were, respectively, released in September 1997, bundled with Windows from 1998 to 2003, and a heading in Microsoft's own documentation for the first of them.

Not established: when it was created, by whom, and on whose request. There is no archived control message for it. Nothing here dates the group, and the dates in this article are the dates of the technologies and of the branch around it, never of the group. No subscriber count, no article count, no thread, no participant and no volume is known or asserted. Whether it was busy, quiet or empty is not settled by anything consulted for this page.

Two inferences are tempting and both should be resisted. The first is that a group with a compressed, three-subject name must have been a low-traffic afterthought; the name may equally record a decision to give one busy subject one room rather than five thinly used ones, which is what the collapse of the French dev groups in August 2000 suggests was the branch's habit. The second is that the absence of a control message means the group was never properly created; 30 of the 103 surviving French names are in the same position, and 382 of the 1,770 names in the hierarchy, which makes the absence a fact about the archive rather than a fact about the group.

Scope and limits

This page is a directory entry and the honest statement of what it rests on is short. For the group itself: two catalogue files distributed by the Internet Systems Consortium, and the absence of an entry in a third. For the branch: 163 archived control-message files under microsoft.public.fr., containing 310 articles, read directly, with their dates, senders and stated reasons taken from the messages themselves and quoted only where quoted exactly. For the technology: the W3C's own dated Recommendations, Notes and Working Drafts, the first and third editions of ECMA-262 read for their definitions of an identifier, an archived copy of Microsoft's own Dynamic HTML documentation, Microsoft's current documentation of the security zones and of the Windows XP Service Pack 2 changes, and standard reference accounts of the browsers, the applications and the worms.

What has not been done: no archive of this group's traffic has been read. No posting is described, no participant named, no thread summarised, no volume estimated. Nothing on this page should be read as a claim about what the group's readers actually asked or answered — including the observation about Outlook Express, which is a statement about how the software was distributed and used generally, not a count of this group's subscribers.

Several adjacent subjects are deliberately left to the pages that own them. The construction and governance of the microsoft.public.* namespace, the recognition programme that identified its most active answerers, and the closure of the news server are on the microsoft.public.* hierarchy page. VBScript, the Windows Script Host and classic ASP are on microsoft.public.scripting.vbscript. The Win32 security model is on microsoft.public.platformsdk.security. The commercial war between the two browsers, and the litigation that followed it, are on alt.netscape.sucks. French-language Usenet, the fr.* and alt.fr.* hierarchies, French terminology policy and the keyboard as a question of language are on alt.fr.comp.os.ms-windows.xp. Each of those pages is linked from the relevant section above. The full list of what this gateway carried is on the all-groups list.

One general caution applies to every number in this article that comes from the consortium's files. Those files describe a namespace, not a network: a name in them means that administrators were offered the name, not that any server carried it, and certainly not that anybody posted. The counts are exact as counts. They are not measurements of activity, and nothing here treats them as such.