Rewriting the History of TCP

A recent statement made in a Cable TV trade publication, from a bandwidth optimization vendor says this about the internet Transmission Control Protocol (TCP):

TCP was designed for Local Area Network (LAN) data transport, not the Wide Area Networks (WANs) used by service providers, and it can consume from 60 to 80 percent of the available bandwidth because of the overhead built into the protocol

This is unfair to the TCP designers, and the history of the internet in general, so let's look at the TCP definition (RFC761) from 1980 to see what it has to say about this. First, in section 1.1 (Motivation):

Very few assumptions are made as to the reliability of the communication protocols below the TCP layer. TCP assumes it can obtain a simple, potentially unreliable datagram service from the lower level protocols. In principle, the TCP should be able to operate above a wide spectrum of communication systems ranging from hard-wired connections to packet-switched or circuit-switched networks.

No LAN bias in those words. Then, from section 2.1, where it discusses the underlying network:

It is assumed here that the networks may be either local networks (e.g., the ETHERNET) or large networks (e.g., the ARPANET), but in any case are based on packet switching technology.

(ARPANET was a wide area network built mostly of leased lines, many of which were running at 56 Kbps). Again, no explicit design centered on LANs.

I think most would agree that TCP's flexible accommodation of a variety of underlying network types, ranging from slow packet radio networks to 10G ethernet is the reason it is still today's transport of choice. Hopefully, the history of TCP's elegant design and tradeoffs will not be lost to a marketing message.

Note that the internet community also recognized the TCP overhead problem for low-speed networks, and came out with a standard method of compressing TCP headers to a few bytes, in 1990.

Posted at 10:38AM Oct 20, 2008 by Peter Schow in Industry  | 

Vintage Univac newsreel from the 1950s

Nice 17-minute video from the 1950s showing the inner-workings of a Univac, complete with VU meters, tanks of mercury, 6-bit characters, "clean desk" programming with flowchart templates, and the obligatory whirring tape drives. In the vocabulary of the 50s, the programmer is assumed to be a "he" even though Grace Hopper was one of Remington's chief programmers at the time.

This video was posted by the folks at the Computer History Museum.

Remington Rand Presents the Univac

Posted at 08:37AM Feb 27, 2008 by Peter Schow in Industry  | 

Free access to ITU-T Standards

It was interesting to see that the ITU has opened up free access to their standards documents.

Twenty years ago, the ITU documents were only available in expensive printed sets that only large corporate libraries could afford. The document sets were color-coded (I remember yellow, red, and blue), issued every four years, and they took up an entire shelf. Given the scarcity and cost of the ITU documents, if you were a network protocol developer in the telco or (now dead) OSI worlds, part of your regular job involved the necessary sleuth work, bribery, and bartering to get the relevant standards you needed. It was never easy, even if you did work for a company that owned a document set.

In stark contrast, the Internet Engineering Task Force RFC documents and drafts were always open, available via FTP. You never had to worry about someone lifting your printed copy of the SMTP (email exchange) protocol spec, for example, while you were away; if that happened, you just downloaded another. Attempting to build upon the very succesful IETF model, Carl Malumud based an entire book on a worldwide quest in the early nineties to have these ITU standards opened up, including a visit to the then-stodgy ITU headquarters in Switzerland. His trips to emerging countries, most still based on dial-up or low speed leased lines at the time revealed that there was great interest in some of the ITU standards, for implementation and experimenting, but they were just too expensive to acquire. With ITU permission, Malamud and Dr. Michael Schwartz even setup their own ITU document retrieval FTP server at the University of Colorado for a trial period.

17 years later, the ITU documents are freely downloadable. In addition to running code and interoperability testing, free document access is one of the reasons why the IETF protocols are dominant today. Perhaps the ITU decision is 20 years too late?

Posted at 04:43PM Sep 25, 2007 by Peter Schow in Industry  |  Comments[1]

Electrifying the Porsche again, 107 years later

With the news that some MIT students are electrifying a Porsche 914, it's interesting to look back to the year 1899 where Ferdinand Porsche himself was working on electric cars in Vienna.

According the the book "Small Wonder - The Amazing Story of the Volkswagen" by Walter Henry Nelson, "electromobiles" were all the rage at the turn of the 19th century in Europe with the added benefit that they were quiet and free of fumes. In 1900, Porsche did build the electric Porsche-Lohner Chaise, capable of traveling 50 miles distance (80 kilometers) at a time. A few years later, he built a combination electric/gasoline car, call the "Mixt", in which the internal combustion engine powered the generators that fed the electric motors in the wheels.

Everything does come around again.

Posted at 05:12PM Jul 01, 2007 by Peter Schow in Industry  | 

Telecommuting in 1860 ?

The recent biography of Andrew Carnegie by David Nasaw has a mentioning of an early telecommuting experience in the USA, pre Civil-War, in the year 1860.

Upon getting promoted to the superintendent of the Pittsburgh section of the Pennsylvania Railroad company at the very young age of 25, Andrew Carnegie returned to Pittsburgh from Altoona in 1860 and settled in the Homewood (now Point Breeze) neighborhood, eight miles away from the dark sooty air of downtown. Running the railroad was a 24 hour job, requiring him to stay in constant contact with his work crews, who were always on the job repairing broken rails. Being an experienced telegraph operator from his teens, Carnegie promptly had a telegraph extension wire installed to his house from the East Liberty railroad office which allowed him to run the business from home.

True telecommuting, 16 years before telephones were invented, and at least 110 years before the term itself was coined!

Posted at 03:02PM Apr 01, 2007 by Peter Schow in Industry  | 

HPC Computing pioneer Neil Lincoln passes away

Eugene Miya reported last week on Usenet that computer architect Neil Lincoln quietly passed away on January 26. If you haven't heard of Neil before, he was a top supercomputer designer of his time and was an architect of the CDC STAR-100 and CYBER 205, and ETA Systems ETA-10 supercomputers. I'm quite surprised that there is no Wikipedia article about him but that will be remedied soon!

Even if you were aware of Neil, some things that you may not have known him:

• He was a seminary drop-out, Vietnam Vet (Marine Corps) pilot, ex-nightclub singer, and former radio DJ


• As an EE student at Cal Berkeley, he and some friends converted a room in an old house into "The Wire Room", known throughout the campus as the place where a computer was being built by those "without the foggiest notion" of how to go about doing it


• In Vietnam (1962-1963), he was responsible for militarizing an IBM 1401 computer on a bed of a heavy truck.


• He very much disliked the wearing of suits and ties, especially at work!


• He did software too: on a contract with CDC, he led the team that developed the SCOPE 2.0 Operating System for the famous CDC 6600 system, and the associated Fortran compiler

He eventually joined the STAR-100 team in 1967, being led by well-known CDC architects Jim Thornton and Seymour Cray. As you probably know from history, Cray disagreed with the vector-intensive direction of the STAR-100 and went on to form his own famous computer company. Jim Thornton also left CDC to start Network Systems (the Hyper Channel folks), eventually acquired by StorageTek (nowadays, known as Sun Microsystems, of course). Neil Lincoln stayed with CDC and in 1975, became the architect of the CYBER 203/205 systems.

Four (vector) pipe CYBER 205 at Florida State University Computing Center, 1988.



During two short contracting gigs at ETA Systems in St. Paul, MN, 1986 and 1987, I remember seeing Neil quite a bit in the evenings. In 1986 his office was in the middle of a busy corridor near the center of the building and his walls were plastered with individual pages of a calendar datebook, showing the countdown to the important "run a job" milestone (this was simply abbreviated "RAJ" everywhere in the company). With new hardware and a brand new OS, running a batch job from start to finish was the confidence test that would exercise the entire OS/hardware path, including the compilers, scheduler, I/O, filesystem, and of course, the CPU. RAJ did run late and ETA was forced to ship system #1 without a full OS. Jobs could still be run on this system but they had to be started in "monitor mode" with no scheduler running. To get Fortran programs running on the ETA at FSU in January 1987, we compiled them on the adjacent CYBER 205, and moved the binaries to the ETA Service Unit with none other than the trusty Kermit communications program, which always seemed to "just work", no matter what system it was running on.

In those days of very expensive supercomputers, new generations of hardware had to be marketed long before they were available, to ensure a decent pipeline of customers at FCS time. In fact, it went farther than this: systems were sold before FCS with contracts that had financial penalties if they were not delivered and accepted by the customer by a certain date. As such, Neil was often asked to talk with incoming customer delegations that were visiting the plant. The evenings, I'm guessing, were his time to get away from meetings and concentrate on the engineering issues of the day. As chief engineer of the ETA-10, he was overseeing new hardware, new cooling (the CPU boards were submerged in liquid nitrogen), and a complex new operating system that was being designed to "host" other OS environments, like Unix. Needless to say, there was not a shortage of engineering challenges in any of these areas.

The ETA Systems software development environment was on Apollo Domain workstations running Apollo's DSEE software engineering environment. Workstations were a shared resource back then, and were not on individual desktops but rather were maintained in various communal areas, like along the windows and in various "war" rooms. You could still do development on a VT220 clone terminal back in your cube, but a workstation was the platform of choice, if you could find an available one. Being away from home with usually no other things to do, and very curious about Apollo's Domain OS/DSEE, I tended to hang around in the evenings when workstations were more available, and it was during this time that I saw Neil a lot. Sometimes he would say "Hi" and one time he stopped by and we talked about the stuff I was working on (Service Unit diagnostics). Even when you could tell the deadline pressure was getting to him, he would always manage a smile. During my second stint there in 1987, when the Lachman and Associates SVR4 UNIX port was up and running on ETA hardware, you could see Neil working at a Unix prompt on one of his systems, looking like he was having fun.

Me in 1988 with ETA Systems ETA-10, Serial Number 1, Florida State University Computing Center, Tallahassee, FL.

Rob Peglar's Saga of ETA Systems from 1990 makes great reading for anyone interested in the ETA Systems story, including his assessment of the business and technical reasons for its eventual failure in 1989. I have some opinions and observations about the complex operating system development effort that was happening at ETA, but that's for another blog.

From a Usenet posting I made back in 2001, I still enjoy reading Neil Lincon's honest assessment of his own computer, the STAR-100, in what did work and what didn't. Computer language enthusiasts may be interested in the APL influence on the STAR-100.

With HPC computing making what seems like a strong comeback, it's important to remember the pioneers of HPC. Neil R. Lincoln is certainly a hall-of-fame member of that club.

Posted at 08:19PM Feb 13, 2007 by Peter Schow in Industry  |  Comments[4]