Envoy, Common Ground and Replica: The Portable Document Formats PDF Beat

A four-way market, not a vacuum

The standard account of PDF runs in a straight line: John Warnock writes a memo, Adobe ships Acrobat, the world adopts it. That story is missing the other three companies in the room. Through the middle 1990s, at least four commercial products competed to be the way a formatted document travelled between machines that did not share fonts, software or operating systems — Adobe's Acrobat and PDF; Envoy, associated with WordPerfect and then Novell; Common Ground, also marketed as Digital Paper, from No Hands Software of Belmont, California; and Replica, from Farallon Computing. Only one of them is still a filename extension anyone recognises.

The launch dating needs care, because the two best-sourced datings do not agree. Adobe's own corporate retrospective places the Acrobat launch in mid-June 1993. A 1996 article in Online and CD-Rom Review, written while all four products were still being compared in print, states that Acrobat, Envoy and Common Ground "all launched commercially within a few months of each other in 1994, as did a format called Farallon Replica that will no longer be marketed from this year". These are reconcilable — announcement and first release are not the same event as volume commercial availability, and Adobe is an interested party writing anniversary marketing twenty-five years after the fact — but they should not be flattened into a single date. This page keeps both visible.

What offices actually did before

It is tempting to describe the pre-PDF era as chaos, with formatted documents arriving broken or not at all. Warnock's Camelot paper, the document usually cited as PDF's origin point, does not describe it that way. It describes two practices that worked, and explains why each was unsatisfactory rather than unusable.

The first was PostScript itself, and its Encapsulated PostScript subset, used as an interchange format — send the print stream rather than the source document. The paper's objection is economic rather than technical: "this solution requires powerful desktop machines and PostScript printers". The archive's page on PostScript covers why that was a real constraint in an office where the expensive machine was shared. That path was also a first-class Internet citizen: application/postscript was, and remains, a registered media type, which is more than can be said for two of Acrobat's three commercial rivals.

The second was fax, and the paper names it verbatim: "The popularity of FAX machines has given us a way to send images around to produce remote paper, but the lack of quality, the high communication bandwidth and the device specific nature of FAX has made the solution less than desirable." That is a description of a working fallback with known costs, not of a gap. Anyone who has watched a contract go out by fax because the recipient's software could not open the file will recognise the reasoning; the practice outlived every product on this page.

The four contenders, as the record supports them

Envoy: an attribution nobody has nailed down

Envoy is the rival that came closest, and it is also the one whose history is least secure. The claims that circulate everywhere — that Tumbleweed Communications introduced it in 1993, that it shipped with WordPerfect Office in March 1994, that versions 1 and 7 were released — all appear to originate from a single encyclopaedia article and propagate outward from there. None of them is corroborated by any primary source located for this page, and one primary source cuts against the origin claim directly.

That source is Tumbleweed's own patent, filed 24 October 1996 and granted 4 August 1998, covering an electronic document delivery system. Its specification refers to "Novell's Envoy" — twice, in the possessive — and never presents Tumbleweed as the format's originator. A company writing its own patent specification has every incentive to claim provenance it holds; this one does not. The conventional chain, WordPerfect Corporation to Novell after the 1994 acquisition to Corel in 1996, is the one the archive treats as most likely, while noting it is not settled.

The same patent is the best contemporaneous evidence of how the market looked from outside by late 1996. Its opening framing is that "Adobe Systems' Acrobat PDF and Novell's Envoy portable document formats have come into widespread use", and the delivery system it describes claims compatibility with both. Common Ground is not mentioned at all. An interested third party, three years into the contest, saw two horses.

Common Ground: registered, and undocumented

Common Ground is usually cast as the open, standards-minded alternative that lost to a proprietary giant. Its own registration document says otherwise, and it is worth quoting because it is the single most load-bearing primary source on this page. On 11 April 1995 No Hands Software registered a media type with IANA: "A Content-Type of 'application/commonground' indicates a document in the Common Ground portable file format, also known as Digital Paper." Under the heading for a published specification, however, the registration offers only "Specification by example: From any version of the Common Ground full viewer, select 'Save As...'", adding that a paper specification "will be available upon request from the address below, beginning in the middle of 1995".

Read plainly: at the moment of its formal Internet registration, the format was undocumented, and the recommended way to learn it was to reverse-engineer output from the vendor's own viewer. Adobe, by contrast, has a documented specification lineage: ISO 32000-1 was fast-tracked from Adobe's own PDF Reference, sixth edition (2006). The date at which Adobe first published a PDF reference was not verified in this pass. Openness is a genuine differentiator in this story — it simply cuts the opposite way from how the story is usually told. Common Ground did get one thing none of the others managed, which was the registration itself; it is the only one of Acrobat's three commercial rivals with a media type in the registry.

Common Ground's most-cited technical idea was the self-viewing document: a file that carried its own viewer, so a recipient needed no software install. The registration records the Macintosh file type APPL and creator CGVM, which is the fingerprint of an executable rather than a document. That is the honest cost of the trick. A file that is a program is a file bound to one platform, so a portable document stopped being portable at the operating-system boundary, and mailing executables around was already becoming a security problem rather than a convenience. The feature is usually presented as something Adobe lacked; it is better read as a trade Adobe declined to make.

Replica, and which Farallon

Replica came from Farallon Computing — and the name needs disambiguating on first encounter, because the company later became Netopia while Farallon survived as a networking product brand. A reader searching the name today lands on Ethernet adapters and LAN hardware, not documents. Farallon of the early 1990s was a Macintosh connectivity company, and the Computer History Museum holds collection material under its name.

Its patent portfolio confirms where its expertise sat: screen image sharing among heterogeneous computers, running local networks over ordinary telephone wiring, network configuration and connection. There is no portable-document patent among them, and a parallel assignee search for No Hands Software returns nothing at all. This is negative evidence, and it needs stating precisely: neither Farallon nor No Hands Software held a portable-document patent at all, so neither was defending one. Whether anyone else's patents mattered here was not tested in this pass.

Replica also went first. The same 1996 review that dates the launches records that Replica "will no longer be marketed from this year" — roughly two years in the market, then withdrawal. No verified account of the reasoning was located for this page, and the archive prefers to leave that blank rather than invent a plausible one.

What the record says actually decided it

Strip out what the evidence does not support and the field narrows quickly. It was not demonstrable file quality, because no public specification for two of the four formats has been located, so nobody outside the vendors could measure them. Only one advantage is left standing: a specification other people could implement against. The usual second half of the answer — installed base — is the part this page could not verify, and the aside below says why.

The registry is only good for a negative: Envoy and Replica never obtained a media type, and Common Ground — the one rival that did — is also the one whose registration admitted it had no specification. Registration was not what separated the winner from the losers. Meanwhile application/postscript was already registered, which meant the incumbent practice — just send the print stream — had the better standing on the network than two of the challengers. Warnock's own retrospective account of the rival document-interchange languages is that they "got horrendously complicated and horrendously hard to implement", which is a claim about implementability rather than about features, and it is the same axis.

The longer consequence is visible elsewhere in the archive. A documented format that anyone could write became the substrate for enterprise document management, for print-production interchange, and eventually for a formal standard when PDF was published as ISO 32000-1 in 2008 — a fate no closed competitor could have reached even if it had survived. The 1990s office context in which this contest was decided, and the desktop-publishing shift that created the incompatibility problem in the first place, are covered on their own pages here.

Timeline, with the dating disputes intact

1. 1991 (Warnock's dating)

   Warnock writes the Camelot paper. He says in a 2018 oral history, "so, in 1991, I wrote this little paper". Secondary accounts split between 1990 and spring 1991; the circulated PDF's 1995 file date is a digitisation artefact and should never be cited as the authoring date.

2. June 1993 (Adobe's dating)

   Adobe's corporate retrospective dates the Acrobat launch to mid-June 1993.

3. 1994 (contemporary review's dating)

   A 1996 review states that Acrobat, Envoy and Common Ground "all launched commercially within a few months of each other in 1994", as did Farallon Replica. On the conventional chronology, Novell's acquisition of WordPerfect — and with it Envoy — falls in the same year.

4. 11 April 1995

   No Hands Software registers application/commonground with IANA — the only one of Acrobat's three commercial rivals to obtain a media type — while stating that the format's specification exists only "by example".

5. 1996

   The same review reports Replica "will no longer be marketed from this year". Tumbleweed's patent application, filed 24 October 1996, names only PDF and Envoy as being in widespread use.

6. 2008

   PDF is published as ISO 32000-1, an outcome available only to a format with a public specification.