Hacker News Reader

I've noticed that the closer I am to historical events, the more wrong reporting of those events tends to be. I have a neighbor who worked at RAND in the 60s and will ask him about this next time we meet.

Worth noting perhaps that many (all?) technical innovations are the result of some underlying technology maturing to the point that it can be applied to a problem. In this case, I bet that nobody liked the fragility and brittleness of circuit switched networking, but in order to make a packet switched network you need small fast computers that are cheap enough to deploy as network nodes. These appeared : minicomputers. The first ARPANet nodes were minicomputers running routing software. In fact the Internet used regular computers as routers into near modern history (IBM RISC machines iirc were deployed at the DS3 upgrade). So PSN is the result of a) people sitting around wishing they could have a PSN, and b) the technology to actually realize that becoming practical. There's no eureka moment.

It’s impossible to know everything that’s happening at the same time. So while ARPANET was the 2nd packet switched network behind NPL a journalist wouldn’t have a clue.

The original conception of Message Blocks was for routing around a damaged network, but the term for Packet Switching was actually a means for multiple users to share a single connection. ARPANET included many ideas from NPL but was also its own thing.

So who invented Packet Switching depends on what parts of it you consider critical.

> I've noticed that the closer I am to historical events, the more wrong reporting of those events tends to be.

That is not just for historical events, but for anything you are familiar with.

Michael Crichton coined the term Gell-Mann Amnesia Effect

https://theportal.wiki/wiki/The_Gell-Mann_Amnesia_Effect

> IBM RISC machines iirc were deployed at the DS3 upgrade

You recall correctly. Info/pictures here[0].

0 - https://www.rcsri.org/collection/nsfnet-t3/

Interesting, thanks for the link. Do you know anything more about the router cards? It says they had intel i960s that ran AIX, but I wonder if it was just the RS/6000 that ran AIX, and the boards ran some custom firmware.

There isn't much published information about these systems to my knowledge. This[0] report from ANS-Net[1], though, indicates to me that AIX was very much directly involved in the forwarding part of the router architecture:

  AIX Kernel Radix Trie problem

  3+ overlapping routes with different masks

    abc.def/8
    abc.def.0/10
    abc.def.0/12

  Fixed in AIX build coming within two weeks

While it would certainly be possible (and is a common router architecture now) to have a central forwarding database and distributed line cards, it doesn't really make any sense to use a kernel trie to store routes for such an architecture, and if the central rs/6000 was switching every packet, it wouldn't - at the time - have been able to support a single t3, let alone the kind of port density described in the original article.

So, I guess in summary, I can't point you to definitive proof that those boards were running AIX, but it seems extremely likely to me.

0 - https://www.ietf.org/proceedings/29/ops/netstat.becker.slide... 1 - A network operator that was subsequently purchased by AOL.

The i960 cards ran an independent AIX kernel as well, housed in a POWER RS6k running AIX.

Not really knowing anything about software portability between different computer architectures or the intricacies of high-speed networking, at first I thought that was kind of strange.

But I checked and it looks like by the time these cards we created, AIX had already been ported to the IBM's ROMP, 370, and POWER processors, plus the 386, so AIX would seem to be fairly portable. I guess it would be easier to port it to the i960, than to do something from scratch. From marcus0x62's link, it looks like very few ( < 100?) of these boards were made.

AIX was for a time IBM's brand name for UNIX environments so it covers some rather distinct products. The version that runs on ROMP is very distinct (one of the earlier commercial microkernels) from the S/370 and i386 version (which, for a time were similar, then there was an ESA/390 version that is not at all), which are both very far removed from the version that continues today on POWER. Sauer was heavily involved in the stuff before POWER and is authoritative on the early history https://notes.technologists.com/notes/2017/03/08/lets-start-.... I can extrapolate a bit more if necessary.

I am unsure of what exact kernel strain wound up on the i960 cards, suffice to say any of them could have been called into this service since a well contained UNIX port is something that can be done by relatively few people.

IBM provided the IBM RT with special cards as the T1 implementation for NSFNet, and then the RS/6000s were used later and especially for higher speed T3 and FDDI nodes.

While the original NSFnet nodes and cards would be rare indeed, IBM commercialized this all as the 6611 Network Processor family.. which ranged from tiny PowerPC-only machines (based on the 7011) to custom chassis complete with the i960 cards. These sold from the early 1990s to mid-late 1990s and were not uncommon. Vanilla RS/6000s were also common in that era as networking multitools and were plenty good at it.

* https://www.ibm.com/common/ssi/ShowDoc.wss?docURL=/common/ss...

* http://ps-2.kev009.com/basil.holloway/ALL%20PDF/sg245000.pdf

* http://www.bitsavers.org/pdf/ibm/6611/

Appreciate the information / links, that's why I like Hacker News. Now, instead of doing something boring, like watch the Olympics, I'm going to read up on some old networking processors.

For those who are interested in that sort of thing, Baran's full 11 publications outlining the design goals and principles of packet-switched networks in January of 1964 are available from RAND as downloadable PDFs:

<https://www.rand.org/about/history/baran.html>

The first document introduces the basic problem:

Let us consider the synthesis of a communication network which will allow several hundred major communications stations to talk with one another after an enemy attack. As a criterion of survivability we elect to use the percentage of stations both surviving the physical attack and remaining in electrical connection with the largest single group of surviving stations. This criterion is chosen as a conservative measure of the ability of the surviving stations to operate together as a coherent entity after the attack. This means that small groups of stations isolated from the single largest group are considered to be ineffective.

<https://www.rand.org/pubs/research_memoranda/RM3420.html>

"Attack" isn't defined, but it's clear that resilience against broad assaults was a key consideration from the very beginning.

> "Attack" isn't defined, but it's clear that resilience against broad assaults was a key consideration from the very beginning.

It was a consideration for Baran's work, but not for ARPAnet:

From chapter two of Wizards:

> Roberts also learned from Scantlebury, for the first time, of the work that had been done by Paul Baran at RAND a few years earlier. When Roberts returned to Washington, he found the RAND reports, which had actually been collecting dust in the Information Processing Techniques Office for months, and studied them. Roberts was designing this experimental network not with survivable communications as his main—or even secondary—concern. Nuclear war scenarios, and command and control issues, weren’t high on Roberts’s agenda. But Baran’s insights into data communications intrigued him nonetheless, and in early 1968 he met with Baran. After that, Baran became something of an informal consultant to the group Roberts assembled to design the network. […]

* https://www.goodreads.com/book/show/281818.Where_Wizards_Sta...

* https://en.wikipedia.org/wiki/Larry_Roberts_(computer_scient...

Influences can be manifold and subtle. What's clear is that in 1964, as packet-switched networks were first being considered, resilience against "attack", however defined, was a key consideration for one significant set of innovators. I suspect some of that logic was incorporated into ARPANET's eventual design, whether the nominal head designer was aware of this or not.

One underappreciated aspect of complex projects is how different goals and intentions may exist simultaneously, and how much active participants even at a high level may not be awarer of this. A friend had a uni professor who'd been part of the Glomar Explorer scientific mission which served as a cover for Project Azoran, the clandestine recovery of a Soviet nuclear submarine. The professor was unaware of the actual mission until well after the mission had concluded, when he read about it in the newspaper.

<https://en.wikipedia.org/wiki/Project_Azorian>

I tend to strongly discount personal-experience denials of covert or secondary mission roles by people otherwise connected with an activity (e.g., company employees, contractors, government workers, contractors, etc.). It's not that these are never, or even mostly incorrect. It's simply that for a large enough project, some secondary goal might well exist without the conscious awareness of many of those involved.

And again, in the case of Baran, we have the receipts.

> but I think we can say that it was _designed_ to survive a nuclear strike

There is a lot more to the design of 'the internet' than the selection of a packet switched protocol. The early boxes were not hardened in any way and were intended to support computer timesharing amongst academic researchers. The DoD's C2 providers themselves rejected the concept of a decentralised packet-switched network because 'it would never work'. Which is why the relevant theoretical papers were sitting on the shelf and available when ARPANET was designed

> but I think we can say that it was _designed_ to survive a nuclear strike

On what basis? What is the distinction between being "created" to survive a nuclear strike, and being "designed" to do so?

> that was one of the reason that packet switching was invented (compared to the traditional, at the time, circuit switching).

Yes, but I don't think it's a relevant one. Baran's papers kinda-sorta-maybe had some influence on ARPANET, but ARPANET mostly got packet-switching (and certainly the term "packet") from Donald Davies. If you look at the actual layout of ARPANET it wasn't very survivable (not much redundancy in the links) [0], compared to Baran's proposal [1]. Internetworking and "the Internet" as we know it came much later and was way beyond the point where Baran had any influence.

[0]: https://commons.wikimedia.org/wiki/Category:ARPANET_maps [1]: https://personalpages.manchester.ac.uk/staff/m.dodge/cyberge...

> I don't think that the Internet was created to survive a nuclear strike, but I think we can say that it was _designed_ to survive a nuclear strike, that was one of the reason that packet switching was invented (compared to the traditional, at the time, circuit switching).

That cannot be said unless you can cite sources saying so. It's been a little while since I read Wizards, but I don't remember any mention of nuclear war until Baran's work at RAND is mentioned, which is a couple of chapters in (IIRC).

For Licklider et al it was all about research, collaboration, and resource sharing.

> The idea of packet switching as a way make a communication network more robust came to Paul Baran a few years earlier, and while it was not the basis of Arpanet, it probably influenced it.

Baran's work was used in things like queuing theory, but work was already underway on ARPAnet before Baran (and Davis in the UK) was roped in. In fact it was Davis that pointed out Baran's work to the ARPAnet folks:

* https://en.wikipedia.org/wiki/Donald_Davies

This is all covered in Where Wizards Say Up Late.

1) Robustness in the face of nuclear disaster was one of the drivers of packet switching.

2) Not the only one, though.

3) It was such a good idea that it grew a life of its own, and no one talked about nuclear war anymore.

1) Only perhaps for Baran. Didn't enter into the equation for, e.g., Davies in the UK.

From the Prologue of Wizards:

> Bob Taylor, the director of a corporate research facility in Silicon Valley, had come to the party for old times sake, but he was also on a personal mission to correct an inaccuracy of long standing. Rumors had persisted for years that the ARPANET had been built to protect national security in the face of a nuclear attack. It was a myth that had gone unchallenged long enough to become widely accepted as fact.

> Taylor had been the young director of the office within the Defense Department’s Advanced Research Projects Agency overseeing computer research, and he was the one who had started theARPANET . The project had embodied the most peaceful intentions—to link computers at scientific laboratories across the country so that researchers might share computer resources. Taylor knew theARPANET and its progeny, the Internet, had nothing to do with supporting or surviving war—never did.Yet he felt fairly alone in carrying that knowledge.

> Lately, the mainstream press had picked up the grim myth of a nuclear survival scenario and had presented it as an established truth. When Time magazine committed the error, Taylor wrote a letter to the editor, but the magazine didn’t print it. The effort to set the record straight was like chasing the wind; Taylor was beginning to feel like a crank.

* https://www.goodreads.com/book/show/281818.Where_Wizards_Sta...

I would think that Taylor of all people who know:

* https://en.wikipedia.org/wiki/Robert_Taylor_(computer_scient...

From chapter two of Wizards:

> Roberts also learned from Scantlebury, for the first time, of the work that had been done by Paul Baran at RAND a few years earlier. When Roberts returned to Washington, he found the RAND reports, which had actually been collecting dust in the Information Processing Techniques Office for months, and studied them. Roberts was designing this experimental network not with survivable communications as his main—or even secondary—concern. Nuclear war scenarios, and command and control issues, weren’t high on Roberts’s agenda. But Baran’s insights into data communications intrigued him nonetheless, and in early 1968 he met with Baran. After that, Baran became something of an informal consultant to the group Roberts assembled to design the network. […]

* https://en.wikipedia.org/wiki/Larry_Roberts_(computer_scient...

The Wizards book came along 20+ years after everything happened, so all Markoff could do was interview a few people.

I said it was ONE of the motivations; not the only one. And Bob Taylor was only one of the people involved; there were lots of others. Why a bureaucracy decides to support something is pretty much impenetrable and lost to history.

> The Wizards book came along 20+ years after everything happened, so all Markoff could do was interview a few people.

Are there any interviews or documentation from ARPA or the early days of ARPAnet that do say it was about nuclear war survival?

Licklider, Taylor, Roberts, Davies in the UK: no one else thinking about things in that way from all the reports and interviews I've seen. The only one that perhaps seems to have had an interest in it seems to have been Baran, and he came in later when the ball was already rolling.

> I said it was ONE of the motivations; not the only one.

Okay. Where are the documents and/or interviews of those involved in the decision-making and/or implementation process stating it was one of the motivations?

I don't know why you're arguing.

Obviously we're talking about a period where documentation is sparse or non-existent, and most of the players are dead. Why is this important to you?

> Obviously we're talking about a period where documentation is sparse or non-existent, and most of the players are dead. Why is this important to you?

Because if you are going to make a claim it'd be useful if you had a citation for it. Because we're as close to the events that occurred as is possible, and it's only going to get 'worse' as time passes, so we should try to get right now.

You say that "all Markoff could do was interview a few people". The phrase "few people" is doing a lot of heavy lifting: two of those "few" were Herzfeld (who ran ARPA at the time) and Taylor (who was in charge of getting ARPAnet going). If they didn't know why ARPA why created ARPAnet then who else would?

If there were/are supporting documents, interview, etc, which show that surviving a nuclear attack was a motivation, they should be shared.

I don't necessarily care what the motivations were, but I'd rather not folks repeating hand-wavy claims that don't seem to have any supporting documentation for. That was what the linked to web page is about in the first: trying to trace and dispel an apparent myth.

Absence of evidence is not evidence of absence.

I have no obligation to do a lot of research for you which, I'm sure, would make no impression on you anyway.

> I have no obligation to do a lot of research for you which, I'm sure, would make no impression on you anyway.

"What can be asserted without evidence can also be dismissed without evidence." — Genghis Khan (or https://en.wikipedia.org/wiki/Hitchens%27s_razor )

OK. I'll just dismiss your claims, too, and we're done.

> OK. I'll just dismiss your claims, too, and we're done.

LOL. I'm the one quoting Wikipedia and books and putting links to references in his posts. You're the one that has not put a single citation in anything that he's written in this sub-thread.

You're making a historical claim and he's disputing it. Is your counterargument is that it was a long time ago so we can just make up whatever we want?

No,the "counter argument" as you put it, is that if you only "know" things that come with a link, you have no value over an LLM

> No,the "counter argument" as you put it, is that if you only "know" things that come with a link, you have no value over an LLM

As opposed to knowing things without evidence and trying to convince people of a position without providing evidence? Is that how history is supposed to be done, without evidence?

As least with a link you can gauge the quality of the reference and its sources.

I think it's also worth noting that the physical infrastructure of certain portions of the circuit-switched AT&T Long Lines network (which also carried the AUTOVON network for the DoD) were designed, or at least attempted to be designed, to survive a nuclear strike. Theoretically traffic would have been manually re-routed around destroyed areas in a post-strike scenario.

There's certain underground bunker sites on the historical L4 transcontinental coaxial cable routes which were designed with equipment mounted on springs, massively thick concrete walls, decontamination showers, and so on. Other mountain top and flat land area microwave point to point relay sites were designed to take a certain amount of overpressure and possibly would have survived nearby smaller nuclear weapon strikes (though obviously not a direct hit).

Much of this pre-dates 1969 and the concept of the ARPANET. If you dig into things like what telecom services/links supported the NIKE missile batteries built around major cities and so forth. And what telecom services/medium to long distance links supported the data feeds from regional radar sites into the SAGE direction centers.

Here's a fairly typical example of the more costly site of site, which you can see the photos of under construction and how much extremely reinforced underground concrete is involved.

https://coldwar-ct.com/Home_Page_S1DO.html

Perhaps corroborating your point, excerpted from Waldrop's 'The Dream Machine':

"Why did ARPA build the network?” Lukasik asks. “There were actually two reasons. One was that the network would be good for computer science.""

...

"But there was also another side to the story, which was that ARPA was a Defense Department agency. And after Eb [Rechtin] came in, defense relevance became the dominant notion. Everybody was writing relevance statements. "

...

"So in that environment, I would have been hard pressed to plow a lot of money into the network just to improve the productivity of the researchers. The rationale just wouldn’t have been strong enough. What was strong enough was this idea that packet switching would be more survivable, more robust under damage to the network. If a plane got shot down, or an artillery barrage hit the command center, the messages would still get through. And in a strategic situation—meaning a nuclear attack—the president could still communicate to the missile fields. So I can assure you, to the extent that I was signing the checks, which I was from nineteen sixty-seven on, I was signing them because that was the need I was convinced of.”

Waldrop, M. Mitchell. The Dream Machine (p. 273). Stripe Press. Kindle Edition.

Yes, quite. This is a very silly article which ignores a lot of the real history in favour of a cutesy top up of early 1990s nostalgia - which was a good 20-30 years after the events that really matter.

Saying "The Internet isn't ARPANET" is ridiculous. Of course it isn't. ARPANET was an academic research project with a mix of defence and open R&D requirements. The Internet is a collection of extra layers of commercial development on top of some of that R&D.

Academic research projects are rarely hardened because the point of the project is to investigate possibilities, not to spend hundreds of billions building a physically bomb-proof network that's useless because the core tech doesn't work.

When the Berlin Wall came down the goals changed, but the core concept of distributed scalable robustness is still very much there today. Of course now we have too many choke points, so it's not as robust as it could be. But if someone cuts a cable packets will still find a longer, slower way around as long as the bandwidth is there.

In the case of all out nuclear war with the USSR the choke points would be the 3000 or so targets that would be hit by Soviet nukes.

In my opinion packet switching is an idea that would have been inevitable once the technology to support it is in place. The real important idea of the Internet is the insight that the envelope should be independent of the data it contains.

I say this because I work with specialized equipment that uses RS-232 serial protocols for communication, and despite many decades of examples available of "the right" way to do it (e.g. OSI model), engineers continue(d) to not understand this lesson and design ad hoc protocols that don't respect this division of concerns and which suffer for it. Even in IP protocols designed to modernize this to wrap the RS-232 serial packets in internet packets, they repeat the same mistake(s). That is, in the midst of having to deal with a more complicated problem for serial protocol to IP protocol conversion because the original format didn't clearly distinguish envelope from data, they compound the problem by mixing levels in the new protocol.

For example, writing IPv4 into the standard for a proprietary an Application or Session level protocol even at the same time as our network is banning IPv4 and requiring IPv6, but the protocol is not even defined for IPv6 addresses. When it should be agnostic to IP level concerns entirely. (The engineers said, while designing the system, "well .1 is going to be this piece of equipment, .2 is going to be the other side, if it's blah blah blah it's .3, etc.")

People keep repeating this mistake at all layers of the stack because they fail to see different layers as managing different sets of concerns and want to treat it as a monolith.

A very recent and high level example of this is how the Matrix protocol was architected. By defining all the message formats in terms of JSON-over-HTTP it makes it difficult to only use part of the protocol and not all of it, and makes it difficult to use it over alternative transports since the assumptions of HTTP idioms are baked all the way down.

This is not true. For instance, https://matrix.org/blog/2021/06/10/low-bandwidth-matrix-an-i... shows how you can (very easily) swap http+json for coap+cbor as an alt transport for Matrix.

This is a great example of what I mean. They had to swap for formats that were designed to be equivalent to HTTP and JSON in order to not have to rewrite large parts of the protocol. And that project was just a prototype that was never productionalized to my knowledge.

> The idea of packet switching as a way make a communication network more robust came to Paul Baran a few years earlier,

I remember back in the days when the internet was first starting to happen, seeing comments about packet switching. The gist of what I recall was that there was lots of dark fiber, and folks like AT&T were trying to prevent it from being used and force slow/expensive packet switching on us, billing us per-packet or something.

it is hard to dig up any article about this however.

When was this? I would have thought the internet was well established before fiber optic cabling was widely used.

AUTODIN preceded it, but ARPANET won out over AUTODIN II.

The first time I heard of it: https://en.wikipedia.org/wiki/Automatic_Digital_Network

For all intents and purposes the Internet was created by and for US military not for public utility, and one of the main reasons for its creation, if not the sole reason, is to be able to withstand large scale nuclear attack. It's not created for remote feeding the cats with the IoT at home while you are away. It's easily the top ten invention of 20th century.

It's not only Internet, FFT also was created to detect illegal nuclear testing and it's included the top ten algorithms of 20th century [1],[2]. There's no shame of admitting the facts.

If not for military purpose Internet will probably never ever see the light of day. The Internet packet switching end to end network precursor and predecessor in France proposed and implemented by Louis Pouzin, famously called the 4th man of the Internet, who coined the word "datagram" the very fundamental concept in packet switching, lost his proposal to other communication technology for France nation wide networks implementation [3].

The original inventors of the Internet proposed the idea to AT&T, the US communication behemoth and monopoly at the time laughed at the packet switching idea saying that it will not work and does not make any economic sense. The only reason Internet created and survived is because of the military applications and as any military application the main objective is to maintain and sustain command and control where Internet is more than capable of. I've once read someone simulated 99% infrastructure demolition (as in nuclear attack) and Internet communication managed to survive intact. Cannot recalled the article now, perhaps this is where ChatGPT becomes handy. The fact that the US military maintained a separate military network (similar to Internet) from the public Internet after the Internet gone public and its popularity increased, provided the hints why it's so important for the US military.

[1] Great algorithms are the poetry of computation:

https://www.andrew.cmu.edu/course/15-355/misc/Top%20Ten%20Al...

[2] The Algorithm That Almost Stopped The Development of Nuclear Weapons:

https://www.iflscience.com/the-algorithm-that-almost-stopped...

[3]Louis Pouzin:

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

> For all intents and purposes the Internet was created by and for US military not for public utility, and one of the main reasons for its creation, if not the sole reason, is to be able to withstand large scale nuclear attack.

And yet Charles M. Herzfeld (who was ARPA director at the time) and Robert Taylor (who was in charge of getting ARPAnet going) say otherwise: it was to link computing centres across the country so that resources could be shared and collaboration would be easier.

If Herzfeld and Taylor didn't know why ARPA created ARPAnet then who else would?

The entire linked to article is about tracing the source of the "able to withstand large scale nuclear attack" original story myth:

> However, the documentation is voluminous and the people who were in the room have all given a consistent story about how it was to build a network for time-sharing of expensive computers and better collaboration.

Interestingly Charles M. Herzfeld once chaired the Nuclear Weapons Council and the Intelligence R&D Council.

According to Wikipedia, apparently "Taylor had convinced ARPA director Charles M. Herzfeld to fund a network project earlier in February 1966, and Herzfeld transferred a million dollars from a ballistic missile defense program to Taylor's budget". This budget probably equivalent to tens of millions USD in today's money. The narrative "to link computing centres across the country so that resources could be shared and collaboration would be easier" using a large chunk of budget from ballistic missile systems that could carry nuclear warheads that were distributed across the country does not make sense to me.

Packet switching was invented by an American and Brit independently IIRC.

Yes, the linked article talks also about Donald Davies, which influenced more directly Arpanet.

But the point about Paul Baran is that for him packed switching was a way to make communications more resilient in face of nuclear bombing (and other things) and he had at least some influence on the birth of the Internet, so if you ask me "what is the relation between the Internet and a nuclear strike" my first answer is "Paul Baran"

> But the point about Paul Baran is that for him packed switching was a way to make communications more resilient in face of nuclear bombing (and other things) and he had at least some influence on the birth of the Internet […]

From chapter two of Wizards:

> Roberts also learned from Scantlebury, for the first time, of the work that had been done by Paul Baran at RAND a few years earlier. When Roberts returned to Washington, he found the RAND reports, which had actually been collecting dust in the Information Processing Techniques Office for months, and studied them. Roberts was designing this experimental network not with survivable communications as his main—or even secondary—concern. Nuclear war scenarios, and command and control issues, weren’t high on Roberts’s agenda. But Baran’s insights into data communications intrigued him nonetheless, and in early 1968 he met with Baran. After that, Baran became something of an informal consultant to the group Roberts assembled to design the network. […]

* https://www.goodreads.com/book/show/281818.Where_Wizards_Sta...

* https://en.wikipedia.org/wiki/Larry_Roberts_(computer_scient...

Baran's influence was perhaps in theory and algorithms, not in intent/use.

For a couple of decades there was a totally independent network stack in the UK (JANET) with its own equivalent of RFCs and so on. History gets written by the victors.

a nuclear attack could be natural selection for the internets evolution honestly.