Talk:ARPANET

TCP/IP Again Called "Foundation." MCI Propaganda?

"The ARPANET was an early packet-switching network and the first network to implement the Internet protocol suite, commonly known as TCP/IP, which became the technical foundation of the Internet." it sez here.

I associate this "the technical foundation" stuff very much with the public relations department of MCI, puffing up that company's own importance after they acquired Vint Cert ("Cerf," obviously, but I love "Cert" as a Freudian slip!) as a director or senior officer. (Somebody please correct me on which it was.)

MCI's pioneering, to my mind, deserves most kudos not for TCP/IP, but for their microwave line between St. Louis and Chicago, which both demonstrated it, and a lot of other good engineering, no doubt, but also cleared out the legalistic underbrush in Washington, and made it clear to the world that the age of copper wire was over. (Some fashion designer at the time was trying to make us all wear green suits, but I have only ever seen two. One was on the MCI lobbyist who visited my office in the Rayburn House Office Building in 1968 or '69, and the other on a Secret Service man I almost bumped into when President Nixon passed me on his way to the Senate Dining Room about that time. They, the suits, not the lobbyist and the security guy, then vanished without a trace, and upscale polyester was the rag trade's next inane try.)

The fact is the Internet was perking along quite nicely, i.e. it had its foundations in place, with all those bang-bang-bang addresses, before the excellent Cerf came along and eased expansion with his development of TCP/IP/.

TCP/IP was useful, and is today necessary, but it's the ceiling of the first floor, or maybe the floor of the mezzanine.

It's not any foundation.

David Lloyd-Jones (talk) 11:52, 26 August 2015 (UTC)

Use of the definite article

Please, it's always 'the' ARPANET. See e.g. the DARPA report "A History of the ARPANET: The First Decade", which always uses the definite article; it never uses the bare term "ARPANET" as a subject, object, etc. I'm not sure how the use of bare "ARPANET" got started (perhaps by someone who wasn't a native English speaker), but it's wrong. Noel (talk)

ARPANET Images

I suppose it doesn't matter much, but there is at least one error in the 1977 ARPANET image in the infobox. Rutgers had a PDP-10 followed by a DEC-20 on the net. They never had a PDP-11; although they later had a VAX-11, not connected to the IMP. O3000 (talk) 12:11, 24 October 2017 (UTC)

That's the Rutgers IMP - but the "Rutgers" is merely a name for the IMP, it was actually owned by DCA. DCA made the decisions on what hosts were connected to each IMP - and one PDP-11 that's connected to that IMP is, according to a copy of the Hosts file from 1977, host 3/46, "NUSC", a PDP-11/40 running ELF, front-ending a Univac 1108. NUSC probably stands for "Naval Underwater Systems Center" or Command. According to the Hosts file, "UPENN" was also connected to that IMP (host 1/46); the file doesn't say what kind of machine, but according to the map, apparently another PDP-11. You'll notice both of them have a little 'v' in a corner; that means they were both 'Very Distant Hosts', a way to connect up over a phone line that allowed the host to be a long way from the IMP. Noel (talk) 00:17, 3 December 2017 (UTC)

Thanks. Seems unlikely. I was head of systems and communications at Rutgers when the IMP was installed. But then, even if my memory was perfect, I'm not an RS.:) O3000 (talk) 14:35, 3 December 2017 (UTC)

Hey, don't take my word for it: you can find a 1977 copy of the Hosts.Txt file at the end of this, with the non-Rutgers hosts on that IMP listed there. You can find confirmation of the meaning of the "v" legend here: the small "v" represents a Very Distant Host (VDH) interface to the IMP. If you want to know more about VDH interfaces, find a later copy of BBN Report 1822. Noel (talk) 00:11, 5 December 2017 (UTC)

ARPANET’s relation to the modern internet

This article could really do a better job of explaining the relation of ARPENET to the modern internet. Some sources describe “the internet” as direct descendent of ARPENET while other suggest it was created as a separate network that operated alongside ARPENET for a short while but using technologies first developed and tried out in mass on ARPENET and that for a time they where operating in parallel before the last bits of ARPENET where shut down. So exactly what is the deal? Was the modern internet essentially what was once ARPENET but now under a different name or did they both exist in parallel at one time, with ARPENET only contributing packet-switching technology to the internet but with both being technically separate networks? Notcharliechaplin (talk) 03:51, 15 February 2019 (UTC)

BBC article

Interesting bit at https://www.bbc.co.uk/news/business-49842681 which includes a couple of early maps: one hand-drawn by Larry Roberts dated (by the BBC) 1969, and another undated but apparently preceding the ones used in this article. These are credited Getty but presumably also exist in the public domain.

The latter map is divided into East Coast (Mitre, Burroughs, Harvard, BBN, MIT and Lincoln), West Coast (Rand, SDC, UCLA, Stanford, Ames, SRI and UCSB) and other (Utah, Illinois, Case and Carnegie) sites, and every site is served by at least two connections. MarkMLl (talk) 07:23, 16 October 2019 (UTC)

Burroughs is rarely mentioned in this context, but at the time they had a significant involvement with military computers and communications hardware. I'm told that

"From it's position on the map, I suspect it was either the Burroughs Research Center in Paoli, Pennsylvania (suburban Philadelphia), or their Great Valley Laboratories (GVL), about four miles away, next door to the Tredyffrin facility where the B8500, B7700, and other large systems originated. The Tredyffrin is still a Unisys facility, although they no longer own it and have leased it for years. The last I knew, they occupied only half of the building.

"I think the location is likely GVL Building #3. That is where the Illiac IV was assembled in the late 1960s and early '70s. I worked in GVL #1 during the summer and fall of 1970 as a new hire, and my apartment-mate was an electrical engineer working on the I/O subsystem of the Illiac IV. I saw the IV several times. I also saw an IMP in the room with it, although my roommate had to tell me what it was. So there was definitely at least the possibility of an ARPANET node in GVL #3 in 1970.

"The military/special projects would have been next door in the GVL buildings. The Tredyffrin building was originally constructed for the B8500 project, so I think it belonged to corporate engineering. After the B8500 product failed, that site became the home of the B7000-series systems. The military/space business got sold off to one of the big U.S. contractor firms a long time ago (probably in the '90s)."

Above from Paul Kimpel in the context of discussion of his B5500 emulator. MarkMLl (talk) 07:14, 18 October 2019 (UTC)

References in film and media

This section was removed, wondered why?

• The removal happened some weeks ago thru an anonymous edit and was followed by a bunch of vandalism. Probably the removal was vandalism too.

Now it is back. Anyway. Please check this section "Scenario", an episode of the U.S. television sitcom Benson (season 6, episode 20—dated February 1985), was the first incidence of a popular TV show directly referencing the Internet or its progenitors. The show includes a scene in which the ARPANET is accessed.

I asume that Riptide was earlier because they are receiving information via a satelite connection in the pilot premiered on January 3, 1984. Unfourtunately I have only found a German version of it (https://www.youtube.com/watch?v=zEEvTWoDdoo). The time codes are 0:56:18 - 0:56:22 but mainly 0:57:18 - 0:57:52.--P. Adamik (talk) 12:51, 13 November 2019 (UTC)

References

Commercial partnerships

I'm not sure DARPA contracting with BBN to add TCP/IP qualifies as a commercial partnership - the Department of Defense contracts with commercial entities for a number of reasons.

If the intent here is to discuss commercial adoption of TCP/IP, as per Internet protocol suite#Adoption, that's the commercial adoption of TCP/IP for non-ARPANET use. As that section indicates, it wasn't just the BBN BSD stack. Guy Harris (talk) 21:49, 5 February 2020 (UTC)

Thanks for the suggestion Guy, I linked to that section in the article. Whizz40 (talk) 09:33, 6 February 2020 (UTC)

Time-sharing of mainframe computers didn't require the ARPANET

You can time-share a computer - mainframe or otherwise - by directly attaching terminals to the computer; that doesn't require packet switching or even circuit switching. You can also have dial-up access, which just requires circuit switching on the plain old telephone service network.

In addition, time-sharing normally referred to conversational access, of the sort that, on the ARPANET, was provided by Telnet. The ARPANET also supported FTP and email.

(And, of the four initial machines on the ARPANET, only the IBM System/360 Model 75 is generally considered a "mainframe". The SDS 940 could be considered a midicomputer, and the SDS Sigma 7 could either be considered a minicomputer or perhaps an early supermini; the PDP-10 might be considered a mainframe, but it was generally thought of as a different type of machine from an IBM mainframe.) Guy Harris (talk) 08:30, 22 April 2020 (UTC)

Agree, thanks for clarifying. Whizz40 (talk) 08:34, 22 April 2020 (UTC)

Always thought of the PDP-10 as a midicomputer. Even the later Dec-20 I didn't think of as a mainframe, although some called it such. SDS-940 definitely mini. I thought of the SDS/Xerox Sigma series as mainframes, as they used the IBM instruction set, like the RCA Spectra series. RAX was a little used timesharing system on the System/360. Pre-system/360, there was Quiktran on the IBM 7040/7044. Of course there were earlier timesharing efforts. I'm getting old. O3000 (talk) 10:41, 22 April 2020 (UTC)

Dubious claim in lede

1st to use TCP/IP? It's my understanding that testing was simultaneous with SATNET, and full adoption was done by others before arpanet? Satnet went over in early 82, ARPNET in Jan 83 (officially, defacto dec 82)...86.142.118.35 (talk) 23:12, 14 July 2020 (UTC)

I'll change the lede to read ARPANET was one of the first to adopt TCP/IP. Whizz40 (talk) 04:03, 15 July 2020 (UTC)

"By 1973, e-mail constituted 75% of the ARPANET traffic" as a dubious claim

Hey all, Even though it is undoubted and actually referenced that e-mail has grown very quickly in ARPANET traffic since its inception, this actual figure seems unreferenced and even considered as dubious in a recent discussion of the History of computing "SIGCIS" mailing list. References 9 and 84 do not actually state that figure at all and should refer to the previous sentence. I am thus witching the references to the previous sentence,removing the dubious claim, replacing it with a similar, but unquantified, phrasing, adding a valid reference to it. Feel free to discuss it here. Alexandre Hocquet (talk) 23:57, 29 July 2020 (UTC)

Quite right. I handled a node and I wouldn't be surprised at the stat. But, I've never seen any actual published stats. O3000 (talk) 00:03, 30 July 2020 (UTC)

Done. Here is a link to the discussion on the SIGCIS mailing list as a ref Alexandre Hocquet (talk) 00:08, 30 July 2020 (UTC)

It seems there is a source for 75%, which has been referenced in Hafner (1996) and reused elsewhere:

• "By 1973, a study “found that three quarters of all traffic on the ARPANET was email”" (Hafner 1996, p. 194). https://pne.people.si.umich.edu/PDF/isis_review.pdf
• "An ARPA study in 1973, a year after network email was introduced to the ARPANET community, found that three-quarters of the traffic over the ARPANET consisted of email messages." https://www.forbes.com/sites/gilpress/2015/01/02/a-very-short-history-of-the-internet-and-the-web-2/?sh=3fad22937a4e
• "The report was commissioned by then ARPA director, Stephen Lukasik, who himself was a big fan of email." https://sites.cs.ucsb.edu/~almeroth/classes/F04.176A/homework1_good_papers/jay-akkad.html
• "By the early 1970's, three-quarters of all traffic on the Arpanet was e-mail." https://archive.nytimes.com/www.nytimes.com/learning/teachers/featured_articles/20011207friday.html

Looks like this can be restored in the article. Whizz40 (talk) 20:38, 30 December 2023 (UTC)

Invention of the Internet

The following statement in the last sentence of the lead seems inaccurate:

"The ARPANET project was formally decommissioned in 1990, after partnerships with the telecommunication and computer industry paved the way for future commercialization of a new world-wide network, known as the Internet."

This implies that the Internet was newly created after ARPANET was decommissioned, which contradicts established facts, including other Wikipedia articles such as History of the Internet. To suggest that the Internet came about only after it was commercialized is obviously false, as this ignores the fact that email, Usenet, IRC, etc. were developed and active much earlier. The actual infrastructure and technologies of the Internet did not magically change after 1990. The Internet was known as the Internet for many years beforehand. In short, what I am saying, is that the Internet in 1990 was not a "new world-wide network." Laval (talk) 08:15, 9 November 2020 (UTC)

No, it did not imply this. The text merely implies that this network was more commonly known as the Internet from thereon, while previously it was in fact more commonly referred to as the ARPANET. The new term gradually became more prevalent after the NSF expanded the ARPANET with new infrastructure circuits that had no connection to military funding.kbrose (talk) 14:25, 10 November 2020 (UTC)

Reworded to read as follows:

"The ARPANET project was formally decommissioned in 1990, after partnerships with the telecommunication and computer industry had paved the way for the adoption of the Internet protocol suite."

Whizz40 (talk) 20:49, 9 November 2020 (UTC)

Excellent wording. Thank you. Laval (talk) 12:18, 10 November 2020 (UTC)

Therefore this wording is in fact incorrect, or at least misleading, and untimely because the decision to use TCP/IP was far from final, despite already in use for a decade. It was a pragmatic result from commercial preference of developing TCP/IP applications and products. Many academic participants and the large network operators where still promoting the use of OSI technologies. PS: After the NSF funding expanded the network, the result really was as if a new network had been created. Rarely anyone among the academic drivers of the expansion wanted to connect to the old ARPANET branch, but wanted to connect to the NSFNet to connect to supercomputing centers and exchange datasets, papers, and communications among researchers who were constantly pushing for new, faster circuits. In terms of networking technologies, including network management, this was indeed a new network. Internet2 is a similar expression of that. kbrose (talk) 14:25, 10 November 2020 (UTC)

That's not what the sources say and it is absolutely untrue that the Internet in 1990 was "new," as discussed above. Laval (talk) 03:07, 11 November 2020 (UTC)

The sentence does not state that the Internet was new in 1990, but that the ARPANET was decommissioned in 1990, AFTER the arrival of commercial partners and build-out.

What sources? Sources need to reflect history correctly. By 1990, the TCP/IP suite was already in use for almost a decade, it didn't need any paving the way anymore. In fact many industry players were trying to pave the way for OSI protocols, which eventually failed because of the head start that TCP/IP already had. In many ways, the Internet expansion of 1986 forward was a new network in terms of infrastructure, and even of software (c.f. DNS, for example) as the old IMPs were phased out with modern routing equipment interconnect by new links that completely superseded the old infrastructure.

Furthermore, it is widely acknowledged that the ARPANET was a PREDECESSOR to the Internet, and not one and the same, and therefore it is well within bounds that the new network was in fact a new entity. In the academic circles, the main users of the network between 1986 and the very early 1990s, nobody talked about ARPANET anymore, but about NSFNet and Internet. Consequently, the ARPANET was retired in short order.

Other Wikipedia articles, such as NSFNET, correctly reflect this history, and pretty much make it clear that a new network came into existence with the NEW BACKBONES of the NSFNET with speeds of 56kb/s, then T1, and T3 circuits, and other regional networks (such as MERIT, and MIDNET,...), and the commercial backbones (IBM/MCI, ANS). kbrose (talk) 16:35, 11 November 2020 (UTC)

I'll drop the word "new", which may suffice for consensus. Whizz40 (talk) 22:43, 11 November 2020 (UTC)

While that might help, some more nuances are important. The DARPA research formulated, specified (in RFCs), developed and tested TCP/IP. International public data networks already existed as commercial services based on X.25, such as the International Packet Switched Service. So private sector expansion and future commercialization was a matter of time, indeed already existed. An internet was inevitable; even if it relied on gateways. Equally, the Internet that we have today, based on the Internet protocol suite (TCP/IP including DNS, etc) was not quite assured at that point, as discussed above. It was the DARPA partnerships with the telecommunication and computer industry in the 1980s that had catalyzed the adoption of TCP/IP. That, along with its use on the NSFNET, at CERN and other benefits of TCP/IP compared with its alternatives, made it the protocol of choice for adoption on existing networks (e.g. JANET) as well as for the private sector expansion and commercialization as it happened. But even then its adoption for the Internet also depended on changes in NSFNET policies and ultimately NSFNET retirement as well as legislative changes in the US. We should be clear on what we attribute to DARPA/ARPANET. Whizz40 (talk) 13:15, 12 November 2020 (UTC)

Epilogue: dropped the words "new" and "future" and merged the two proposed sentences to read as follows:

"The ARPANET project was formally decommissioned in 1990, after DARPA partnerships with the telecommunication and computer industry had paved the way for the widespread adoption of the Internet protocol suite as part of the private sector expansion and commercialization of a world-wide network, known as the Internet."

-- Whizz40 (talk) 18:59, 27 December 2020 (UTC)

Two new sources for ARPANET history

For editors looking to improve or update this article and seeking reliable sources for ARPANET history, I've recently learned of two historical documents from or about DARPA that have been made available online:

• A History of the ARPANET: The First Decade - published in April 1981 and prepared for DARPA by BBN.
• DARPA Technical Accomplishments - A Historical Review of Selected DARPA Projects, Volume 1 - published in February 1990 by the Institute for Defense Analyses. The ARPANET text begins on page 20-1 which is page 243 of the PDF. (Note that this document covers a wide range of other DARPA projects and so may be useful for improving the history of those Wikipedia articles, too.)

I don't currently have time myself to see if these sources can help improve this article, but I thought I would leave them here in case other editors do have the time. - Dyork (talk) 12:09, 25 October 2021 (UTC)