Search This Blog

Tuesday, July 24, 2012

Ars Technica = Internet History FAIL

From Wikipedia
In a burst of Administration-inspired foolhardiness Ars Technica has decided to pick apart a small WSJ article on the "history of the internet" to show that it was, of course, all done by the government.

Xerox PARC did develop the underlying technology for what we know today as the internet - but both articles leave out the facts and details and gloss over the actual truth of the matter.

The ARPANET was a creature of the US Defense Department's Advance Research Project Agency.  It was invented as an experiment in digital packet-switching networks.  It used something called the Imp-16, a specialized computer, that was connected to other Imp-16's in other locations via satellite or leased phone line.  The "customers" (as you can see from the chart above) were primarily universities and research groups, e.g., Xerox PARC.  The systems connected were primarily Dec-10/20's, PDP-11, IBM mainframes and CDC mainframes.

Unfortunately for Ars this technology was nothing like the "internet" though there are some software elements of it that still exist in name only today.

I used the Arpanet in the early 1980's when I worked at a company called Unilogic, Ltd.  Unilogic sold a software package called Scribe developed at CMU.  CMU was on the Arpanet.  As part of customer support for Scribe there was an Arapnet email.

Arpanet allowed students involved in computer research to communicate via email (mail), log into remote computers (telnet) and transfer files (ftp).  While these applications survive in name today (and they work in much as they did then) the systems on which they ran basically no longer exist except in museums.

But the similarities ended there - the sites were physically hard wired.  Imp-16's required university research money to purchase, defense department blessing to use, and serious mainframes to connect them to.  Significant budgets for leased lines were also required.  Speeds were limited as was access.

At Unilogic you would dial into a terminal server (probably a PDP-11) that had a rack of 1200 baud modems connected to phone lines.  The PDP-11 would act as a front end for a terminal (typically a hardwired computer terminal like DEC VT52) for the PDP-10/20 for various installations around the CMU campus.  This technology was the forerunner of the internet ISP-style modem dial-up and PPP protocols to come later (this was as I recall developed by DEC and enhanced by CMU - private entities both).

Ethernet - developed at PARC - is a hardware protocol that revolves around a shared connection resource and collision detection.  An RF signal is sent over a coaxial cable.  Multiple devices are connected to the cable and at any time any one of them can start sending a message.  Each device also listens to what's being sent to see if another device overwrites it.  If another device does it waits a short,  random amount of time and tries again.

Ethernet would languish mostly inside Xerox and at a few universities for a decade after its invention: the PC era's beginning saw Novell's networking system became the dominant player inside businesses (even ahead of Microsoft).

In the mean time (circa 1980 or so) Tom Truscott invented the Usenet.  Access to the ARPANET was limited to a small number of players involved in defense contracting - this left other universities out of the picture.

Truscott seized upon the idea of the ARPANET and figured out how to make it work with UUCP (Unix-to-Unix copy program).  UUCP was part of the standard Bell Labs (another private institution) unix distribution and it allowed a system administrator to set up late-night unattended copies of files between systems over modem connections.  Truscott figured out how to use this to support email.

While Ethernet languished Usenet flourished.

In the mid 1970's the notion of the TCP/IP protocol was developed (as describe by Ars).  However, this was a software protocol designed to run on packet-switched networks (ARPANET as an example).   And while today's Internet uses TCP/IP the protocol runs on ethernet which is not a packet switching protocol.

In the 1970's Bill Joy at the University of California Berkeley developed what became known as BSD: Berkeley Standard Distribution.  This was his version of Unix that incorporated much software developed by his group including an implementation of TCP/IP.

TCP/IP was a complex undertaking because of how packets and datagrams are handled.  Basically nodes in a TCP/IP network need to handle "partial" packets as data is moved from point to point and, as far as I know, the Berkeley BSD TPC/IP version was the first written for Unix which was accessible to the public at large.

Again, outside universities and private companies this languished throughout the 1980's.

Also during the 1980's SGML (Standard GML) became an ISO standard.  SGML was a precursor to HTML and XML developed from IBM's GML.  SGML was developed as a means for creating documentation that a variety of systems could view, i.e., a standardized way to indicating bold, italic, paragraphs, and so on that could be displayed on any SGML viewer.

(Mary LaPlante, someone who I worked with at Unilogic, Ltd. for many years became head of the industry SGML group.)

SGML's primary issue was that a viewer was very complex and not easy to implement.

There were also many companies at CMU during this time that built software and tools that used hypertext links - this was a common idea at the time.  Some with SGML - others with proprietary versions.  Many ran as private business - most often involved in corporate training.

This left SGML as a mostly defense/aerospace standard for documenting things like airplanes and weapons systems.

So the "Internet" of Berners-Lee was comprised of technical elements that had all been developed a decade or more beforehand.  Berners-Lee assembled them into a new configuration that became popular.

Of course, things like BSD, ethernet hardware, and all the rest had been worked on for a decade by students and industry.  Bell Labs, the inventor of Unix (a private company) seeded the US University system with Unix so people like Bill Joy could create BSD (which still survives in various forms today - the Mac software I am using to write this has a BSD heritage).

Xerox tried and failed to make money with the Xerox Star office (Alto/Ethernet/Dover laser printer).

Again, it donated technology to Universities (no doubt for tax write-offs) so that students would learn what was done and take their ideas to industry.  (CMU had an ethernet interface in its coke machine in the computer science building which allowed students to query the state of what drinks were available and cold from anywhere on the ARPANET.)

To waive your hands (as did Ars) and dismiss decades of hard work by probably tens of thousands of computer science and EE-types to take these technologies and turn them into successful industrial tools is simply beyond comprehension.

In the days of BSD parents still paid to send their kids to college out of their own pockets - is that government funding?  I doubt it...

Bell Labs probably did the most in terms of releasing Unix into the world - Bell Labs was created by the Bell Telephone company - an "evil" big corporation.  Without this contribution Universities would have had to develop these technologies independently and somehow create a standard - an endeavor that would have added significant time and effort to the process I think.

Universities spun off thousands of "high tech" companies to build everything from ethernet cards (for VAX computers such as used at Berkeley) to SGML viewers.  Most privately funded by business and personal capitol.

Did big government play a role in all of this?  Of course.

(Though mostly through the Defense Department - a fact which seems to bother no one.)

But so did state and local universities, private companaies like Bell, incubator and start-up venture capitalists, private industry, and virtually everyone else.  (For example, Sun (now Oracle) was one of the first to seriously prescribe to "network computing.")

The ARPANET is one of many clear ancestors of today's internet - but only a distant ancestor.  But so is Tom Truscott who found a way to escape the "clutches" of the government controlled monopoly of the ARPANET and build the Usenet for "the rest of us."  So is Bell Labs.

In the 1970's DARPA's job was to do just what it did with the ARPANET - advance technology for US defense.  The ARPANET was the inspirational source of a flurry of activity over many decades by many private individuals to create today's internet.

To argue that the government "did it all" is to argue that Defense funding should be increased to do more of the same.

Yet the defense budget is cut.

As a college student in the mid-1970's things like the ARPANET were inspirations - not to accept "as is" but to view as something to better.  And there were many other inspirations: Gene Amdahl, Dennis Ritchie, Ken Thompson, Per Brinch Hansen all gave talks at the school I attended.  These were designed to inspire students to do more than what had been done before.

I have founded a dozen companies over the last several decades - I have employed hundreds of people - all through this inspiration.

But I myself did all the work and, when things didn't go well, no one from the government came to save me.

Which is the most telling I think.

While you can say "the government did all this" or "none of this could have been done without the government" the fact is that the government never cared what I did as long as I paid taxes and did not break the law.

It never helped in any way.

And in the mid-1970's you could only read about things like the ARPANET in private publications like IEEE journals.

To try and steal this computer science legacy from probably millions of people over three decades is truly a crime and a FAIL by Ars Technica.

[ There is no time to include many, many other personal friends, collaborators, and acquaintances from the last thirty years who I know of that contributed in some way or other.  Some are now dead, some are now retired.  But this Ars article is literally a theft of my heritage as a "high tech" businessman. ]


  1. Good post: thanks for sharing!

  2. When was the defense budget cut?

  3. Its been cut and will continue to be cut over the next decade as part of the debt agreement (provided Congress does nothing).