Articles and media on the early history of the web, and WWW.app service

[Rhetorica]

One of the features I'd (eventually) like to have on nextcommunity.net is a collection of historical museum galleries that can provide gentle introductions to NeXT newbies—in particular sections on the NeXT computer's relevance to the naissance of various industries and technologies. Of course the biggest of these is the World Wide Web, so we need a place for that.

As I don't have the time to sit down and belt out a twenty-page deep-dive on the history right now, I'd like to just gather up some of the sources we can find on WWW.app, Mosaic, Lynx, OmniWeb, and other NeXT browsers. (Paging @stepleton!)

The provocation is actually unrelated to this recent interview with Sir Tim (paywall bypass), although it's a good read if you're in the small-internet camp. It spares only half a paragraph for Berners-Lee's actual work on WorldWideWeb.app, mentioning none of it by name, which feels to me like a missed opportunity. (It also annoyingly says he invented it "in" 1989, when the earliest copyright dates we can find are 1990.)

Incidentally it repeats something I'd heard elsewhere—that as of January 1993, when the first pre-release Mosaic build was made available, there were "[only] around 50 websites." Elsewhere I've heard the claim that there were 50 websites just at CERN on this date. This could be easily tested and/or falsified by finding any URLs linked on Usenet from around this time.

Mosaic's first pre-release in January '93 is the key point where the web starts to escape containment—it spends most of the year only available for Unix workstations (and, somehow, Linux) before receiving Windows and Mac versions of the 1.0 release in November '93.

For a brief moment Windows 3.1 users also have access to a browser called Cello, which appears in June 1993 and is discontinued a year later. This Mozilla blog article suggests Cello originated the convention of blue-colored hyperlinks on the Web, but it also lists WorldWideWeb.app as being a product of 1987, before the NeXT computer was even announced, so I'm inclined to take it with a grain of salt. (Maybe Elise Blanchard confused it with CERN's prior internal SGML hypertext projects—but the takeaway about underlining links could not have been that old.)

Lynx originates in 1992 as a Gopher-only browser before gaining rudimentary HTTP support with the 2.0 release, which was sometime in 1993, though I'm not sure of the exact date. The story about the creators of the Gopher protocol seeking licensing fees certainly seems intimately tied to the genesis of the web.

And, of course, no conversation about preserving early web materials would be complete without a link to CERN's abandoned browser port of WWW, which was rendered non-functional by—you guessed it—modern security practices. Could we fix it? Should we fix it? Would it be better to just provide a disk image for InfiniteMac that has a real copy of WWW.app inside it? 

Problematically, WWW.app (and OmniWeb 1.0) uses a version of HTTP that has retroactively been termed "HTTP/0.9": no headers, and no all-important Host field. The browser only sends "GET <path>" and the server only replies with the content. (Not sure how POST worked—todo.) Allegedly there are modules in Apache that still support this, but it means these browsers are very hard to demonstrate unless a transparent proxy can be set up to intercept their HTTP requests.

There are vanishingly few servers that will honor such requests—the only one we've stumbled upon is floodgap.com. If anyone figures out how to configure Apache to handle them properly, I'd be all ears!

[stepleton]

I am paged!

But I think it may be a case of mistaken identity: while I have some keen interests in NeXT history, they focus mainly on the distant origins of the NeXTstep operating system itself in the SPICE and Accent project(s) at Carnegie Mellon University. These involved the PERQ computer, an early graphical workstation (to a rough first approximation, a Xerox Alto clone) of which I have a specimen currently doing a fine job this minute of being a space heater as it runs a graphical demo.

I hope to explore Accent more closely someday, but right at the moment I am investigating an unrelated and delightfully unusual PERQ operating system called Flex. This OS is as dedicated to "procedures" (compiled functions, basically) as NeXT was dedicated to objects.

There is a hypertext connection of sorts: the Flex user interface is basically a hypertext text editor, though literally none of the extensive online documentation (which is itself hypertextual) uses the term or calls attention to this arrangement as unique or noteworthy. Source code is edited in the same editor (with the primary systems programming language being ALGOL 68), with module inclusion also taking the form of hyperlinks.

Incidentally, hyperlinks are called "cartouches" after the ancient Egyptian orthographic concept and are much sturdier than HTML hrefs, being durable handles to whatever item they reference (an item referred to by a cartouche can't shake that reference: the cartouche itself must be deleted). In fact, the OS doesn't really present a filesystem-like interface per se: it's up to you to keep cartouches pointing to your files, etc. in "directory documents" that you maintain, or to find some other organisation scheme.

Flex was developed by a UK Ministry of Defence laboratory (RSRE Malvern), which distributed it to at least a few universities in the UK as well (on stacks of 8" floppy disks). I'm unaware of it leaving British shores during the time it was under active development, and I don't have any reason to think that it influenced the development of the web or web browsers in any meaningful way.

There were a few other early hypertext systems explored on PERQ, some listed here. I haven't seen any claim that these had any substantive influence on the web either, although unrelatedly one of them did enjoy a trial deployment on the USS Carl Vinson (CVN-70) (an American aircraft carrier) involving around 30 PERQs. Given my own experience of the machines' durability I wonder how well they fared in a marine environment.

Both the ZOG hypertext system (the one on the Carl Vinson) and Flex were designed to take advantage of network infrastructure; I think it's likely that Flex cartouches could reach out over the ethernet cable, but I'm not certain of it.

In any case, if you don't mind being a bit fanciful about it, you can extend the history of the NeXT and hypertext even back to its direct ancestor systems.

[ZombiePhysicist]

WWW.app came out around 90 if I recall. I remember getting the first version and how unimpressive it was, mostly because there was maybe 3 at most sites you could see, and their contents were not great. Gopher, particularly graphical versions on NeXT, plus local mounted FTP software was the big thing around the time WWW first came out on NeXT.

First implementation I saw of hyperlinks was at SRI in the mother of all demos. 

They had quite a system there, but it was a singular time share computer, so all those links, were very much local, but maybe across user spaces. Would be interesting to see if any of those early versions spanned more than one computer. Not sure how far the original SRI system expanded.

Also, it's tough to believe that the xerox star didn't copy some version of the SRI hyperlinked system and if so, you got to wonder if any of those multiple alto/star systems let you link across different machines on the network, even if they shared the same storage mechanisms.

[Rhetorica]

Also, it's tough to believe that the xerox star didn't copy some version of the SRI hyperlinked system and if so, you got to wonder if any of those multiple alto/star systems let you link across different machines on the network, even if they shared the same storage mechanisms.

Well, you may remember that everyone felt that Xerox was sitting on their hands because they'd created a monster that they knew would (eventually) end the paper age. A certain Steve Jobs was among those who were mad at them for refusing to develop the potential of what they'd created!

To my knowledge, the closest thing they ever got to real on-screen document navigation was perhaps Smalltalk's hierarchical browser, which looks a lot like Miller columns without the scrollbars—a far cry from links embedded in files. Anything that couldn't be printed or photocopied was anathema to their business interests. I'm sure they told themselves that no one wanted hypertext—they thought they were making digital printing presses, not book readers, so who would buy a five-figure machine to read a document?

Anyway, perhaps the "first" instance of NeXT hypertext (hyperTeXT?) is right there in Webster version 1.0 from NextStep 0.8: double-clicking any word in a definition performs a search for it, which is good enough for my standards!

[Protocol 7]

Thanks for the info @stepleton. I need to investigate PERQ at some stage. My contribution to early hypertext info would be HyperTIES. Something I've wanted to get up and running for some time. Don Hopkins has most/all of the source code available on his site (which I dutifully pilfered) but I have no idea how to work with it.

I keep meaning to get in touch with him for some help but haven't as yet. Partly because I was waiting for a clean copy of NeWS to surface (which finally happened recently) as I was having all sorts of issues with the copy in circulation. Some are likely to be emulation-related, but also I get the feeling NeWS is more at home on Sun3-era machines and as yet there's no decent emulation available (tme doesn't count. It might be all we have but honestly it's a mess). I'm not sure if it would run on SPARC but it would be great if I could make it happen. There's only a few potato-quality images of it in action and I'd love to rectify that.

[ZombiePhysicist]

Also, it's tough to believe that the xerox star didn't copy some version of the SRI hyperlinked system and if so, you got to wonder if any of those multiple alto/star systems let you link across different machines on the network, even if they shared the same storage mechanisms.

Well, you may remember that everyone felt that Xerox was sitting on their hands because they'd created a monster that they knew would (eventually) end the paper age. A certain Steve Jobs was among those who were mad at them for refusing to develop the potential of what they'd created!

To my knowledge, the closest thing they ever got to real on-screen document navigation was perhaps Smalltalk's hierarchical browser, which looks a lot like Miller columns without the scrollbars—a far cry from links embedded in files. Anything that couldn't be printed or photocopied was anathema to their business interests. I'm sure they told themselves that no one wanted hypertext—they thought they were making digital printing presses, not book readers, so who would buy a five-figure machine to read a document?

Anyway, perhaps the "first" instance of NeXT hypertext (hyperTeXT?) is right there in Webster version 1.0 from NextStep 0.8: double-clicking any word in a definition performs a search for it, which is good enough for my standards!

I believe SRI's 68 was significantly earlier, and if you watch the full demo (which i have many times because im a dork) you see a true hyperlink system was there. Also, there are several innovations he had at the time that few people mention, that are still very advanced for today, and that have not fully been optimized as much as they had to then. They had this kind of just-in-time byte op code they had to develop for on the fly processing (some parts of which are kind of in java just in time engines, but not really anything as amazing as they demo'd in 68. Sadly I don't know off hand where they discuss it in the video. One day I should go through and time stamp the insane number of innovations they had there (like live teleconferencing and dimmed screen overlays in 1968).

But the hypertext system, in that demo, imo, works exactly like modern browsers. You click a linked word, and you 'zoom' to the linked contents. It was very innovative and insane for the time.

I agree, the smalltalk environment seemed to have the first GUI multi column browsers. But maybe the miller column browser was first. I hear that name often, but never investigated its source/origins, which I imagine are very interesting.

[Rhetorica]

I believe SRI's 68 was significantly earlier, and if you watch the full demo (which i have many times because im a dork) you see a true hyperlink system was there. Also, there are several innovations he had at the time that few people mention, that are still very advanced for today, and that have not fully been optimized as much as they had to then. They had this kind of just-in-time byte op code they had to develop for on the fly processing (some parts of which are kind of in java just in time engines, but not really anything as amazing as they demo'd in 68. Sadly I don't know off hand where they discuss it in the video. One day I should go through and time stamp the insane number of innovations they had there (like live teleconferencing and dimmed screen overlays in 1968).

Here's one you can't timestamp: this video is often called "The Mother Of All Demos" for a very good reason—it was the first live demonstration of interactive software.

I was obsessed with it college! It was a big influence on my decision to start working on a generic wiki framework, which I still use for most of my sites. I also went so far as to program a simple chording keyboard simulator. My naive guess at the key layout (that the keys were just bits indexing a 32-bit table, A = 1, E = 5) is proven correct if you find the spot in the middle of the video where he's supposed to explain the keyset—even though the actual footage is missing, the output of pressing each key is shown on screen.

But the hypertext system, in that demo, imo, works exactly like modern browsers. You click a linked word, and you 'zoom' to the linked contents. It was very innovative and insane for the time.

Technically it was just part of a very small crowd in this respect; the Hypertext Editing System beat it by a year, and evidently even had a 'back' button. But there are definitely a lot of other things in NLS that are solid gold, including collapsing bullet points in a hierarchical tree.

It is important to remember that until (very) recently, PARC and SRI were competitors. They were located quite close to each other physically but they led their own siloed existences, where Xerox kept the former from doing anything antithetical to the printing business. Interestingly, the Hypertext Editing System seemed to have the best of both worlds, being designed for editing (implying a goal of paper output) despite supporting links during document production...

I agree, the smalltalk environment seemed to have the first GUI multi column browsers. But maybe the miller column browser was first. I hear that name often, but never investigated its source/origins, which I imagine are very interesting.

Wikipedia confirmed my suspicion—that Mark S. Miller's eponymous columns most likely postdate Smalltalk, although there's a [citation needed] attached to it.

[stepleton]

Here's one you can't timestamp: this video is often called "The Mother Of All Demos" for a very good reason—it was the first live demonstration of interactive software.

The MOAD was an incredible demo packed with firsts, and what I'm about to say doesn't take away any of that, but: can this really be so? Or could there be some qualifiers missing (like "at a conference" or similar)?

Interactive computing is not a lot older than MOAD but it does predate it by some years... Dartmouth BASIC was available to students in June 1964; APL first went interactive in 1966 and was available for sale in '68. Prior to these, Ivan Sutherland had showed off Sketchpad in 1963, although that video is a recording of course. There are probably other examples.... Surely some of these systems were demonstrated live in some form or another, at least to students in classrooms or to prospective buyers?

[Rhetorica]

The MOAD was an incredible demo packed with firsts, and what I'm about to say doesn't take away any of that, but: can this really be so? Or could there be some qualifiers missing (like "at a conference" or similar)?

Interactive computing is not a lot older than MOAD but it does predate it by some years... Dartmouth BASIC was available to students in June 1964; APL first went interactive in 1966 and was available for sale in '68. Prior to these, Ivan Sutherland had showed off Sketchpad in 1963, although that video is a recording of course. There are probably other examples.... Surely some of these systems were demonstrated live in some form or another, at least to students in classrooms or to prospective buyers?

The qualifier is indeed "live, at a conference." The first system with interactive software was probably Whirlwind I, a vacuum tube flight simulator built in 1951, and some descriptions of command-line interactive systems from the 1960s cite the designers wanting the sort of interactivity and control they had on Whirlwind as an inspiration. (I think CTSS and OS/360 were both called such.) The NORAD SAGE radar system's first computers were derived from abandoned blueprints for a Whirlwind II system, and the PDP-1 was also an iteration on Whirlwind-inspired designs. (The list of firsts for that particular machine is pretty long.)

[Rhetorica]

A fairly robust history of Gopher:

The perspective isn't quite complete—it doesn't do a good job of painting a picture of what the 'net was like before Gopher. There's no discussion of BBSes, FTP, or Usenet, which may leave the audience overestimating Gopher's importance in 1992. (It also doesn't acknowledge that WWW.app was released a few months before Gopher 1.0, choosing instead to only focus on Mosaic's success in late '93, but, eh.)

[ZombiePhysicist]

I remember we had some cool gui gooher clients. I know we had ftp clients that mounted in the workspace but there may have been one gopher client that did too.