A map of the entire internet as of May 1973 | Hacker News Reader

A map of the entire internet as of May 1973

twitter.com

235 pointsdoener50 comments

Hi there—original tweeter here.

Some context for the picture—my father, Paul, was the business manager for the computer science department at Carnegie Mellon at the time, and is a fabulous record-keeper, so he kept all sorts of interesting things from that era.

He was also one of the founders of Three Rivers Computer as well as PERQ, which were tech transfer spin-outs from the CS dept in the seventies.

Thank you for posting this, and engaging here on HN. Do you think this collection of interesting things would be a good candidate for digitizing and getting up on the Internet Archive?

Quite possibly—most likely the PERQ and Three Rivers Computer stuff is the hardest to find, though I don't know how much of it is already out there.

bitsavers.org are probably the best bet to have a chat with about this sort of stuff.

Thank you!

This version from 1974 includes the first international link to UCL in London. As no-one had budget for a dedicated transatlantic link, it was piggybacked on a link via NORSAR in Norway that was intended for seismic monitoring of nuclear tests.

http://personalpages.manchester.ac.uk/staff/m.dodge/cybergeo...

...being a bit of a pedant (and Norwegian!) I'd like to point out that the first international link was precisely the one to NORSAR at Kjeller; the link over the North Sea to UCL was the second. :)

(Kjeller/NORSAR was hooked up in June 1973 using a blistering fast 9.6kbps link, if memory serves)

The history of this link is quite interesting:

http://nrg.cs.ucl.ac.uk/internet-history.html

Anyway, I think we can agree to share the glory; although the first international link was to the TIP at Kjeller, the first end host was the PDP-9 in London, which itself acted as a relay to the University of London CDC.

I think you are right; however, I will have to dig into my old copies of the Norwegian engineering monthly Teknisk Ukeblad; I read a very interesting feature there a few years ago, interviewing a couple of then NORSAR employees who admitted that quite a lot of tinkering and exploring resulted once the link was there - as you'd expect, given the nature of boffins...

Edit - I'd like to dig out the old feature not to dispute your suggestion; only to share what was a very entertaining read. :-)

If Norway indeed had the first endpoint, they surely would have opted to use Norsk Data computers rather than PDPs at the time, including coding up an IP ^H^H^H NCP stack for the SINTRAN OS (or its predecessor).

From the looks of it, the first NORSAR computers hooked up on a permanent basis were two IBM mainframes.

Also, the Norwegian Defense Research Institute (FFI) utilized a teletype against the TIP. (Source: A couple of Norwegian-language webpages.)

You're probably right. Btw I didn't want to point out flaws in what you said, but b/c I have fond memories of ND. Out of boredom on a gig a couple years ago I even did a clone of the sintran shell, or rather its unique trie-like command completion over hyphens.

Since you seem interested in norwegian IT history, around 1988 ND had a dept in Kiel/North Germany where I did an apprenticeship and then some freelance work. They did Pascal compilers in cooperation with CAU (Kiel Uni) there, and a commercial system for public libraries.

Very recently, I heard ND also provided computers for controlling parts of the legacy particle accelerators at DESY Hamburg (which is about 3mins from my place).

Thank you; that is most interesting. I haven't had much hands-on experience with ND equipment, only significant exposure I've had to Norwegian-ish computer efforts were working on a Tiki 100 for a couple of years - a mid-eighties Z80 box running a CP/M clone OS, cleverly enough called KP/M (Later renamed to Tiko.) Quite a different beast from the ND minis!

It would make sense that ND had deliveries to DESY - when I was in university (2000-ish), the CS department still had lots of posters on display showing ND equipment installed at CERN; apparently, someone knew someone who knew someone who needed a real-time setup for particle accelerators - and a couple of racks of NORDs running SINTRAN fit the bill nicely.

By the way, the most exotic ND product I've heard of was a 28-bit mini made to circumvent COCOM rules - I can only guess that it specifically forbade export of 32-bit machines to COCOM countries. So, what does an enterprising supplier do? Neuter a 32-bit machine into 28-bit and sell them like hotcakes... :) (I'll have to look that one up one day; I only heard of it over beers (multiple!) at a faculty Christmas party fifteen years ago.)

From the support staff at the Kiel facility I heard that to upgrade from a 500 series to a 5000 series system, all they had to do was bring a pair of pliers and cut of a wire on the board (and install a new OS). They were having trouble explaining this to customers ...

IBM, around 1990 at the latest, also had their AS/400 (now iSeries) and their RS/6000 (pSeries) machines integrated down to CMOS level.

Sintran was great. It could do a couple tricks that others couldn't, such as having processes being frozen and restarted (you can do essentially the same with Unix cores but Sintran had it as a user-level feature).

What I really liked about Sintran was its command shell. Its system utilities were named like "list-files", "list-print-jobs", and so on. Then you could abbreviate it to just "li-fi" or "l-f" and the shell would pick the right command, provided the abbreviation was unambiguous. This was a great way to discover commands, while also being able to use short forms for quick entry. And it worked better (and more stylish, too) than Unix wildcards on a German keyboard for touch typists since the hyphen character could be entered without Shift.

But around when I worked for ND, they already had a System V port running in the lab and prepared for the demise of Sintran (I was actually tasked to develop an offline entry system running on Xenix, among other things). Also, the compilers department they had in Kiel was actually the "B-Team", in that ND's own language creation was the PLANC programming language, and of course developed by the elks themselves, as we called them. I believe they only originally did develop the compilers in Kiel because its main use was said commercial bibliography system at the same site. Nevertheless, when Sintran was seen less strategic for ND, for a short time the compiler guys in Kiel hoped they could gain importance within ND because they had also a C compiler running.

I really appreciate you taking the time to share those Sintran tidbits; I've taken the liberty of copying your comments into a text file so that I'll be able to revisit them later. Thanks!

(On occasion, I head up to my Alma Mater - where a group of geeks* maintain a rather eclectic collection of computer hardware; with any luck, someone with fond memories of Sintran times will be around next time I show up!)

*) www.pvv.org

Looks like a scan of a German book or paper. Any idea which one?

One interesting aside is you can see from that even this early in the evolution of what's still a very sparse network, how dense the resources are in California, presaging the eventual concentration of the "internet business" there. Sensitive dependence on initial conditions and such.

It's also interesting to note that the DC area is the #2 (arguably tied with Boston) given how relatively inconsequential that community is now to what we'd consider the tech sector. That was a slow process of course, given that at one point AOL dominated the consumer internet from its offices near Tysons Corner.

"It's also interesting to note that the DC area is the #2 (arguably tied with Boston) given how relatively inconsequential that community is now to what we'd consider the tech sector."

Remember, back then it was ARPAnet. ARPA was/is the Defense Departments Advanced Research Projects Agency. If you think the defence department then or now as not part of the tech sector you would be underestimating the investment in technology research made by the DOD.

But of course. California's nascent technology industry was also created by Federal dollars, just look at Ames, Rand, etc, on the left side of the map. The interesting question I was pondering is why the transition from public to private/consumer sector happened in one place but not nearly so much in the other.

I think at a certain point the question becomes very similar as asking why you got a 6 in a dice roll instead of a 3. You can definitely explain it using Newtonian physics but at a high enough level I'm not sure it really matters.

If you look at maps of the modern internnet, DC has an extremely large collection of international cable terminals. a lot of traffic goes throught it.

That was because Tysons Corner was (and still is) a major hub of the Internet. Initially because of the DOD and military contractors, but a big chunk of Amazon's cloud is located in data centers there and nearby.

This is the area in question:

https://en.wikipedia.org/wiki/Dulles_Technology_Corridor

Several of my professors at the University of Maryland, in the late 70s, had connections to the institutions that had IMPs, like NBS (now NIST), the Pentagon, Mitre, etc. If you asked nicely, they would give you the dialup numbers for the IMP/TIP so you could play on the ARPANET. I had an account at MIT-AI that I accessed that way.

For a deeper dive into the history of ARPANET cartography, check out this paper (many more maps inside!): Fidler, B., and M. Currie. “The Production and Interpretation of ARPANET Maps.” IEEE Annals of the History of Computing 37, no. 1 (January 2015): 44–55. doi:10.1109/MAHC.2015.16. http://ieeexplore.ieee.org/document/7057559/?arnumber=705755...

For a geographic representation, try:

* ARPANET Maps || http://som.csudh.edu/fac/lpress/history/arpamaps/

As you can read there, these maps come from "Heart, F., McKenzie, A., McQuillian, J., and Walden, D., ARPANET Completion Report". This text can be obtained from http://walden-family.com/bbn/arpanet-completion-report.pdf

EDIT: For a less deep link go to http://walden-family.com/bbn/

Wikipedia has a similar image from 1977 on the PDP-10 page. https://en.wikipedia.org/wiki/PDP-10

Make a high quality scan and blow it up so we can all print out our very own internet map!

No problem: https://www.dropbox.com/s/2efes525sd9q5yg/1973%20ARPAnet%20P...

(I'm assuming that doing this doesn't get my dropbox account suspended for some reason...)

Very nice! This'll look great hung over the microcomputer I've scavanged and fixed up (and hopefully one day my collection of microcomputers I will have scavanged and fixed up).

See also RFC 4.

https://tools.ietf.org/html/rfc4

  7  Test messages between UCSB-SRI 11/15/69

   7a  Network configuration

           SRI .
              |  .
              |   .
              |     .
              |       .
              |         .
              ------------
           UCLA           UCSB

Which modern protocols even existed back then? IP? TCP? FTP?

FTP 1971, RFC 114

TCP & IP, 1974, A Protocol for Packet Network Intercommunication, https://web.archive.org/web/20160304150203/http://ece.ut.ac....

What protocol did they used on the lowest level. E.g., what modulation scheme did they use? Did it run over the phone network? Or did they have their own cables?

NCP, the Network Control Program, was the predecessor of TCP/IP. [1]

Host addresses were only 8 bits (zero was reserved), so there could only be up to 255 hosts on the ARPANET.

NCP was a simplex protocol, not full duplex like TCP/IP, so each port=>port connection transmitted data in only one direction, and applications would use an even/odd pair of ports for two-way communication.

The parity of the ports at each end of the connection was required to be different: you couldn't connect an even port to another even port, or an odd port to another odd port. "Homosocketuality" was strictly forbidden by internet protocols! NCP's mandatory "heterosocketuality" was called the "Anita Bryant feature". [2]

[1] https://en.wikipedia.org/wiki/Network_Control_Program

[2] https://news.ycombinator.com/item?id=12422813

> The parity of the ports at each end of the connection was required to be different: you couldn't connect an even port to another even port, or an odd port to another odd port. "Homosocketuality" was strictly forbidden by internet protocols! NCP's mandatory "heterosocketuality" was called the "Anita Bryant feature". [2] > ... > [2] https://news.ycombinator.com/item?id=12422813

Well, that's a whole distasteful part of computer history I was blissfully unaware of before now. As someone fairly young (in my early 20s), it's sometimes hard to remember that things like this are very recent history, so I guess it helps to have some perspective every now and then.

TCP/IP wasn't the official transport protocol until 1983. Before that was NCP (https://en.wikipedia.org/wiki/Network_Control_Program). I got on the ARPAnet in 1982, and remember the transition day.

FTP and Telnet were around by 1974. So was an early mail protocol (described in RFC 524, https://www.rfc-editor.org/rfc/rfc524.txt) which used FTP as mail transfer agent. The first SMTP-like mail protocol was probably MTP in 1980 (RFC 722, https://www.rfc-editor.org/rfc/rfc772.txt).

Go read some RFCs -- they're great sources of early Internet history! Start here: http://www.rfc-editor.org/rfc-index.html

As noted in other replies, the ARPANET predates IP and TCP. Although some Internet applications, like Telnet and FTP, were first implemented on the ARPANET, I don't consider it to be the Internet in any meaningful way. The Internet was conceived as a network of networks, which directly affected the design of TCP/IP. Thus, the Internet proper began with the transition to IP. ARPANET was a test bed, to see if packet switched networking would actually work. It did. So well, in fact, that the ARPANET was used for actual work, until it became obvious that something better was needed to replace it. That replacement is the modern Internet.

I enjoyed this book about the development of the ARPANET: https://www.amazon.com/Where-Wizards-Stay-Up-Late/dp/0684832...

Surprised that Berkeley got online later than several others UCs and LBL.

What do IMP and TIP stand for here?

IMP must be a Interface Message Processor [1], an Arpanet equivalent to our routers. And TIP must be Terminal Interface Processor [2], not really sure of their usage but basically they would allow remote sessions on the network, my guess is that they were some early forms of modern modems.

[1] https://en.wikipedia.org/wiki/Interface_Message_Processor

[2] https://en.wikipedia.org/wiki/Tip

I went down this rabbit hole this morning myself. There's a nice document at Bitsavers (http://www.bitsavers.org/pdf/bbn/imp/BBN1822_Jan1976.pdf) that has details about the two different IMP models, and the TIP and Pluribus IMP (an interesting multiprocessor IMP that I'd never heard about prior to today).

The TIP was the "Terminal IMP" (per the linked document) that provided connections to 64 terminals directly to the network through the IMP (with no intervening host).

Basically, they were routers.

The ARPANet Completion Report is Fascinating; it includes maps of the US with the ARPANet nodes indicated as the network grew. As well as additional logical diagrams of the network. http://walden-family.com/bbn/arpanet-completion-report.pdf

A very similar map from 1972 is included in RFC432: [PDF] https://tools.ietf.org/pdf/rfc432.pdf

It's kind of amazing how fast the internet evolved. Did the original people who were hooking these boxes up ever imagine what it was eventually going to be?

Does anyone know what some of the non-obvious locations are? I'm not sure what RADC, ETAC, and a few of the others are.

RADC was Rome Air Development Center (from memory at the time; I got on the network in the fall of '72 as a frosh at HARV-10), a big funder of ARPA-related research.

I'm pretty sure RADC were the funding sources for much of the CS research going on at the Harvard CRCT [Center for Research in Computing Technology] in those days. My advisor, Tom Cheatham (RIP), was the head of the CRCT and the PI for many of the funded research programs. (Great man.)

All from memory, the less obvious names:

MIT: LCS, AI, Arch. Machine group, Multics, etc.

Harvard: CRCT with a few machines (PDP-10, -1, etc.)

BBN: contractor that developed the original ARPANET technology

Aberdeen and Belvoir: I believe Army testing grounds

NBS: National Bureau of Standards

MITRE (Corp)

Lincoln (Labs)

LBL (Lawrence Berkeley Labs)

Xerox (MAXC was PARC's home-made PDP-10 clone)

SDC (Software Dev Corp)

We had fun back in the "open days" when most systems had open guest accounts, telnetting from machine to machine to see how many "miles" we could go before the whole thing broke down.

(Fun factoid: back in the early days of email, the mail delivery mechanism was anonymous FTP to append to your recipient's mailbox file.)

Eventually (I'm vaguely remembering by 1973-74) there were enough troubles with vandals that all the open accounts were turned off. (I remember watching a TTY where Harv-10 was logging all typing from guest users, see what mischief they were up to.)

"Those were the days, my friend, we thought they'd never end..."

One of my old networks lecturers used to love trotting out this on a slide, pretty sure he had the same map in his attic.

Reply on news.ycombinator.com