5 ms·

There are other aspects of the origin story that seem to be myth. For instance, the idea that the Internet was designed to route around damage, and could also t

There are other aspects of the origin story that seem to be myth. For instance, the idea that the Internet was designed to route around damage, and could also treat censorship as a form of damage, and therefore it was resistant to censorship. I once spent a long weekend reading through the early RFCs. What I found was the obvious recognition of the fact that AT&T, which at the time had a monopoly on telephone services in the USA, was circuit-switched, whereas IP was being designed to be packet-switched, and clearly a packet-switched technology could be more decentralized than a circuit-switched technology. But that is the only shred of evidence I could find in favor of something like an ideology of decentralization. I could not find anything explicit in the design documents of the 1970s and 1980s. As near as I can tell, the more grandiose claims of decentralization don't show up till the late 1990s, and those claims are not especially made by the technologists themselves (with the exception of Bob Metcalfe). It's mostly techno-utopians and libertarians who begin to make the bolder claims about how the Internet is fundamentally decentralized and resistant to censorship. People like Kevin Kelly, writing in Wired Magazine in the 1990s, made the most utopian claims about the decentralization of the Internet. I've looked, and I've had trouble finding evidence of bold claims of decentralization, being made in the 1970s or 1980s, except in the very obvious case case that a packet-switched network is less necessarily centralized than a circuit-switched network. The myth gets built in the 1990s, and is then retro-imposed on the 1970s and 1980s.

throwaway193948 4y ago
maybe all the arpanet design docs are not public?
- kragen 4y ago
  Well, a lot of the early RFCs have been lost.
  - dredmorbius 4y ago
    Interesting, I wasn't aware of that. Checking manually, I found many of the low-numbered RFCs are present, though there are some higher in range which turn up 404. E.g., RFC 1 is online: https://datatracker.ietf.org/doc/html/rfc1 https://datatracker.ietf.org/doc/html/rfc1 But here's a list of missing HTML pages from the first 1024 RFCs: 8, 9, 14, 26, 51, 92, 159, 201, 220, 244, 248, 257, 258, 259, 260, 261, 262, 272, 275, 277, 279, 284, 337, 341, 358, 375, 380, 383, 397, 418, 424, 427, 428, 444, 465, 481, 484, 502, 507, 517, 530, 536, 540, 541, 554, 558, 564, 572, 575, 583, 598, 605, 639, 641, 646, 648, 649, 650, 664, 665, 668, 670, 673, 676, 682, 693, 709, 710, 711, 715, 723, and 853. For anyone who wants to verify: wget --spider --force-html $( for i in {1..1024}; do echo "https://datatracker.ietf.org/doc/html/rfc${i}"; done ) 2>&1 | egrep '(https.*rfc|awaiting response)' | grep -B1 '404 Not Found' | grep -v '404 Not Found' Above RFC-2000 or so, the pattern of omissions starts to look more deliberate, suggesting reserved values: 8, 9, 14, 26, 51, 92, 159, 201, 220, 244, 248, 257, 258, 259, 260, 261, 262, 272, 275, 277, 279, 284, 337, 341, 358, 375, 380, 383, 397, 418, 424, 427, 428, 444, 465, 481, 484, 502, 507, 517, 530, 536, 540, 541, 554, 558, 564, 572, 575, 583, 598, 605, 639, 641, 646, 648, 649, 650, 664, 665, 668, 670, 673, 676, 682, 693, 709, 710, 711, 715, 723, 853, 1061, 1182, 1260, 1839, 1840, 3100, 3200, 3223, 3328, 3333, 3350, 3399, 3400, 3500, 3699, 3799, 3800, 3889, 3899, 3900, 3907, 3908, 3999, 4000, 4099, 4100, 4199, 4200, 4232, 4299, 4300, 4399, 4400, 4499, 4500, 4599, 4600, 4637, 4658, and 4699.
orev 4y ago
You won’t find anyone saying that censorship is the type of damage to be routed around, in a serious engineering way, because that’s more of a colloquialism than a true design principle. Asserting that it was almost seems like a straw man. The concept of routing around censorship is more of a behavior by people using the applications that the Internet has enabled. There’s no RFC for that.
- schoen 4y ago
  Specifically, "the net interprets censorship as damage and routes around it" is a statement of John Gilmore and originally referred to Usenet (not the Internet). In its most literal sense, it refers to the fact that Usenet tried to achieve reliability by replicating a complete set of recent articles by a flood fill-like algorithm, regardless of the specific path that they took to reach an individual node. Therefore, if you deleted an individual article in one place, it would still propagate elsewhere, even if it took a longer route to reach a given destination. It also has a more figurative interpretation in which net users try to maintain connectivity and the free flow of information, and try to find workarounds when either is disrupted.
  - tiahura 4y ago
    TIL Thanks
mlyle 4y ago
> the idea that the Internet was designed to route around damage See, e.g. Baran's writings from 1964, about decentralization, fault tolerance, and survivability of packet switched networks and directly lead to the ARPANet and internet: https://www.rand.org/pubs/research_memoranda/RM3420.html https://www.rand.org/pubs/research_memoranda/RM3420.html > shred of evidence I could find in favor of something like an ideology of decentralization As to decentralization among many providers-- this only really comes into play in 1989, when BGP emerged, and the NSFNet began to connect regional networks and ISPs and internet exchanges began to flourish-- though you can find Baran writing in the late 1970's about the connection of multiple networks and ensuring that single providers or colluding groups of providers are not too powerful. But it was a goal from the beginning. E.g. Pouzin, 1973: "Actually, this last approach boils down to the construction of another network, and ultimate at that. ... This is not to say that a general internetwork agreement will never happen, but that it will happen gradually... This study attempts to put forward a realistic scheme allowing point to point message transfer across several independent packet switching networks. This requires a common agreement in formatting messages, but constraints are minimized while preserving implementation freedom and efficiency." Of course, many questions remained-- protocol translation vs. common protocol? How to do this all, without dynamic routing? What about the regulatory landscape, etc. We lucked out into a pretty good outcome, overall. > could also treat censorship as a form of damage This is a feature of the culture of the 'net it existed in the 80's and 90's based on experience with Usenet (which indeed does kind of route-around-censorship in many cases, but also engendered a group of users who culturally subverted attempts at censorship), and not a feature of the internet protocols themselves.
bane 4y ago
One of my university professors helped start CSNet which ran as an alternative to ARPANet for a few years over X.25 equipment. There's a complicated history but it sorta ran until the late 80s by itself until the Internet kind of swallowed everything. My impression from my professor is that these academic precursors to the Internet were really intended to be centrally controlled by their various committees (as academic efforts usually are). The notion that these packet switched networks would eventually just sort of be run in a mostly decentralized way didn't seem to have occurred to anybody.
msla 4y ago
The myth was that the ARPANET was designed to survive nuclear war. Wikipedia sets this one straight, anyway: https://en.wikipedia.org/wiki/ARPANET https://en.wikipedia.org/wiki/ARPANET > The ARPANET was not started to create a Command and Control System that would survive a nuclear attack, as many now claim. To build such a system was, clearly, a major military need, but it was not ARPA's mission to do this; in fact, we would have been severely criticized had we tried. Rather, the ARPANET came out of our frustration that there were only a limited number of large, powerful research computers in the country, and that many research investigators, who should have access to them, were geographically separated from them. Citation: Charles Herzfeld, ARPA Director (1965–1967) as quoted here: https://arstechnica.com/information-technology/2019/10/50-years-ago-today-the-internet-was-born-sort-of/ https://arstechnica.com/information-technology/2019/10/50-ye...
- Animats 4y ago
  MILNET actually was. It was 5-connected - each node had at least 5 connections to other nodes. Same technology as ARPANET, but DoD-only.
- KineticLensman 4y ago
  Don’t have refs to hand but IIRC the concept of packet switching was developed theoretically as a candidate for a military network but was deemed impractical by the Pentagon’s telecoms providers. However, the theory papers were picked up again by the early Arpaner designers
  - mlyle 4y ago
    https://www.rand.org/pubs/research_memoranda/RM3420.html https://www.rand.org/pubs/research_memoranda/RM3420.html (1964) "We will soon be living in an era in which we cannot guarantee survivability of any single point. However, we can still design systems in which system destruction requires the enemy to pay the price of destroying n of n stations. If n is made sufficiently large, it can be shown that highly survivable system structures can be built—even in the thermonuclear era. "
Animats 4y ago
For instance, the idea that the Internet was designed to route around damage, and could also treat censorship as a form of damage, and therefore it was resistant to censorship. Yes. That comment was actually made about USENET, not the ARPAnet. Because USENET really did that. When two USENET nodes connected, they compared lists of message IDs, and each one requested any messages the other had, but they didn't. The result was distributed flooding - everybody eventually got everything in any group to which they subscribed. If someone censored the messages going across a link, that would prevent the message traveling over that link. But if any other possible path existed, two nodes would eventually compare messages lists, note the missing ID, and transmit the missing message. So censorship was only possible if you controlled all possible connections between two networks. This came up at Stanford in the 1980s, when some IT administrator (Ralph Gorin?) tried censoring rec.humor.funny. But there was one professor (McCarthy?) who had his own USENET node both on the Stanford network and connecting to the outside. It didn't have much capacity, but as long as there was even one low-bandwidth path, any censored messages would eventually take the slow path and arrive. So censorship was futile. But so was spam filtering.
Animats 4y ago
A packet-switched network is less necessarily centralized than a circuit-switched network. This is not actually true. The AT&T Long Lines network was originally very decentralized. When it was first fully automated, there was no central control point. There were 10 regional control centers, but they were not essential to most calls. Eventually there was central control, out of Bedminster, Pennsylvania. Originally all it did was collect data and send out new routing tables every few minutes. It didn't process any calls. Previously the routing tables had been fixed, changed maybe once a month. Calling pattern changes on holidays were a serious cause of overload. AT&T used to run ads asking people to make their Xmas calls earlier in the day. There's "A History of Science and Engineering in the Bell System"[1], written by the old-timers at the AT&T breakup to record how all the old stuff worked. A general understanding of how the more advanced electromechanical switching systems worked is useful in designing high-rel systems. Number 4 and Number 5 crossbar systems had system reliability much higher than component reliability. In the entire history of the Bell System, no electromechanical central office was ever out of service for more than 30 minutes, other than due to a natural disaster or major fire. That record has not been maintained in the computer era. [1] https://www.amazon.com/History-Engineering-Science-Bell-System/dp/0932764061 https://www.amazon.com/History-Engineering-Science-Bell-Syst...
- dredmorbius 4y ago
  A distinction though is that circut-switched loses connectivity when a circuit is broken. Packet-switched does not. You see some packet loss, there may be increased latency or reduced bandwidth, but the packets can still be routed without session loss so long as some valid routing exists. (For TCP --- I'd tell you the UDP joke but you might not get it ...). To that extent, the packet-switched network is more resilient, and this was a conscious design consideration, as can be seen in Paul Baran's original packet-switched networking documents: Simulation results are shown to indicate that highly efficient routing can be performed by local control without the necessity for any central—and therefore vulnerable—control point. The ten-part series starts here: https://www.rand.org/pubs/research_memoranda/RM3420.html https://www.rand.org/pubs/research_memoranda/RM3420.html The series as a whole is desribed and linked here: https://www.rand.org/about/history/baran.html https://www.rand.org/about/history/baran.html
kragen 4y ago
When John Gilmore said in the early 01990s, "The Net interprets censorship as damage and routes around it," he was talking about USENET, not the internet. USENET is a peer-to-peer gossip network, like Bitcoin. John Gilmore is a techno-utopian, a libertarian, and a technologist; he was Sun employee #5, he designed BOOTP (the basis of DHCP), and he wrote substantial parts of GCC and GDB. This saying of his was quoted in Time Magazine in 01993 https://web.archive.org/web/20210408023213/https://kirste.userpage.fu-berlin.de/outerspace/internet-article.html https://web.archive.org/web/20210408023213/https://kirste.us... which seems to be a bit confused about the relationship between the internet and USENET. I don't know how far back it dates, but I don't think he made it up for the article. The claim is, strictly speaking, true of gossip networks like USENET. If node A talks to nodes B and C, which also talk to each other and also to another node D, and node A utters a message that node B chooses not to distribute, the situation from D's perspective is precisely the same as if node B were temporarily down. Node C will still get the message (directly from A, rather than from B) and node D will still get it from C. You can find a lot of technologists who are also techno-utopians and libertarians making grandiose claims about the decentralizing influence of the internet in the cypherpunks archives from 01992 and 01993. At the time, the list was hosted on Hoptoad, John Gilmore's personal machine, in his house. I think there are a lot of ways that the internet architecture is less centralized than the PSTN that aren't related to packet-switching versus circuit-switching. Regular users aren't allowed to run traceroute on the PSTN (though some phone phreaks used to do that sort of thing, and MPLS means traceroute means less than it used to). The PSTN doesn't have a public whois database. It won't delegate even a class C of 256 phone numbers to your PBX for local use (though that was becoming difficult to get on the internet too by the mid-01990s), much less 2^48 phone numbers. You can't get a PSTN AS number so you can multihome your PBX to two different long-distance providers. Historically there was no mapping from logical names to phone numbers, so to call someone on the phone you had to dial an area code and exchange that identified their local phone company and its exchange, so user-visible addresses were inherently owned by local regulated monopoly providers, not by users themselves. Moreover, ideologically the deployment of new applications on the PSTN was centrally controlled and administered by the phone company, and they would run on the "switches" in the buildings without windows, while on the internet any random person could write and run an experimental server on their node, at least if their sysadmin was cool with it. Until the Carterfone decision of 01968 it was illegal in the US to plug even third-party telephones into the PSTN, and direct-connect modems didn't become legal until 01975. After that, lots of innovative applications of the PSTN were actually deployed in a decentralized way, such as answering machines, fax machines, and BBSes, but they were very limited in the features they could provide. By comparison, the early internet was wildly decentralized. Packet-switching does mean that a single computer can directly talk to an arbitrarily large number of other computers at once, instead of just however many circuits its connection supports. But a graph of degree three can already be a robustly cyclic connected graph, so three circuits per node would be enough. Of course, the phone company only provided most users with one at a time, with thousands of milliseconds of latency to establish a new circuit. So it wasn't so much that we imposed decentralization retroactively on the work of the 01970s and 01980s. It was more that we observed that it had resulted.
nl 4y ago
"Decentralization" has many meanings and often organizational and technical decentralization are conflated. The decentralization talked about in 1990's Wired was both because the technology wasn't controlled by a single company (unlike competitors like AOL or MSN).

throwaway193948 4y ago

maybe all the arpanet design docs are not public?

kragen 4y ago
Well, a lot of the early RFCs have been lost.
- dredmorbius 4y ago
  Interesting, I wasn't aware of that. Checking manually, I found many of the low-numbered RFCs are present, though there are some higher in range which turn up 404. E.g., RFC 1 is online: https://datatracker.ietf.org/doc/html/rfc1 https://datatracker.ietf.org/doc/html/rfc1 But here's a list of missing HTML pages from the first 1024 RFCs: 8, 9, 14, 26, 51, 92, 159, 201, 220, 244, 248, 257, 258, 259, 260, 261, 262, 272, 275, 277, 279, 284, 337, 341, 358, 375, 380, 383, 397, 418, 424, 427, 428, 444, 465, 481, 484, 502, 507, 517, 530, 536, 540, 541, 554, 558, 564, 572, 575, 583, 598, 605, 639, 641, 646, 648, 649, 650, 664, 665, 668, 670, 673, 676, 682, 693, 709, 710, 711, 715, 723, and 853. For anyone who wants to verify: wget --spider --force-html $( for i in {1..1024}; do echo "https://datatracker.ietf.org/doc/html/rfc${i}"; done ) 2>&1 | egrep '(https.*rfc|awaiting response)' | grep -B1 '404 Not Found' | grep -v '404 Not Found' Above RFC-2000 or so, the pattern of omissions starts to look more deliberate, suggesting reserved values: 8, 9, 14, 26, 51, 92, 159, 201, 220, 244, 248, 257, 258, 259, 260, 261, 262, 272, 275, 277, 279, 284, 337, 341, 358, 375, 380, 383, 397, 418, 424, 427, 428, 444, 465, 481, 484, 502, 507, 517, 530, 536, 540, 541, 554, 558, 564, 572, 575, 583, 598, 605, 639, 641, 646, 648, 649, 650, 664, 665, 668, 670, 673, 676, 682, 693, 709, 710, 711, 715, 723, 853, 1061, 1182, 1260, 1839, 1840, 3100, 3200, 3223, 3328, 3333, 3350, 3399, 3400, 3500, 3699, 3799, 3800, 3889, 3899, 3900, 3907, 3908, 3999, 4000, 4099, 4100, 4199, 4200, 4232, 4299, 4300, 4399, 4400, 4499, 4500, 4599, 4600, 4637, 4658, and 4699.

orev 4y ago

You won’t find anyone saying that censorship is the type of damage to be routed around, in a serious engineering way, because that’s more of a colloquialism than a true design principle. Asserting that it was almost seems like a straw man. The concept of routing around censorship is more of a behavior by people using the applications that the Internet has enabled. There’s no RFC for that.

schoen 4y ago
Specifically, "the net interprets censorship as damage and routes around it" is a statement of John Gilmore and originally referred to Usenet (not the Internet). In its most literal sense, it refers to the fact that Usenet tried to achieve reliability by replicating a complete set of recent articles by a flood fill-like algorithm, regardless of the specific path that they took to reach an individual node. Therefore, if you deleted an individual article in one place, it would still propagate elsewhere, even if it took a longer route to reach a given destination. It also has a more figurative interpretation in which net users try to maintain connectivity and the free flow of information, and try to find workarounds when either is disrupted.
- tiahura 4y ago
  TIL Thanks

mlyle 4y ago

> the idea that the Internet was designed to route around damage See, e.g. Baran's writings from 1964, about decentralization, fault tolerance, and survivability of packet switched networks and directly lead to the ARPANet and internet: https://www.rand.org/pubs/research_memoranda/RM3420.html https://www.rand.org/pubs/research_memoranda/RM3420.html > shred of evidence I could find in favor of something like an ideology of decentralization As to decentralization among many providers-- this only really comes into play in 1989, when BGP emerged, and the NSFNet began to connect regional networks and ISPs and internet exchanges began to flourish-- though you can find Baran writing in the late 1970's about the connection of multiple networks and ensuring that single providers or colluding groups of providers are not too powerful. But it was a goal from the beginning. E.g. Pouzin, 1973: "Actually, this last approach boils down to the construction of another network, and ultimate at that. ... This is not to say that a general internetwork agreement will never happen, but that it will happen gradually... This study attempts to put forward a realistic scheme allowing point to point message transfer across several independent packet switching networks. This requires a common agreement in formatting messages, but constraints are minimized while preserving implementation freedom and efficiency." Of course, many questions remained-- protocol translation vs. common protocol? How to do this all, without dynamic routing? What about the regulatory landscape, etc. We lucked out into a pretty good outcome, overall. > could also treat censorship as a form of damage This is a feature of the culture of the 'net it existed in the 80's and 90's based on experience with Usenet (which indeed does kind of route-around-censorship in many cases, but also engendered a group of users who culturally subverted attempts at censorship), and not a feature of the internet protocols themselves.

bane 4y ago

One of my university professors helped start CSNet which ran as an alternative to ARPANet for a few years over X.25 equipment. There's a complicated history but it sorta ran until the late 80s by itself until the Internet kind of swallowed everything. My impression from my professor is that these academic precursors to the Internet were really intended to be centrally controlled by their various committees (as academic efforts usually are). The notion that these packet switched networks would eventually just sort of be run in a mostly decentralized way didn't seem to have occurred to anybody.

msla 4y ago

The myth was that the ARPANET was designed to survive nuclear war. Wikipedia sets this one straight, anyway: https://en.wikipedia.org/wiki/ARPANET https://en.wikipedia.org/wiki/ARPANET > The ARPANET was not started to create a Command and Control System that would survive a nuclear attack, as many now claim. To build such a system was, clearly, a major military need, but it was not ARPA's mission to do this; in fact, we would have been severely criticized had we tried. Rather, the ARPANET came out of our frustration that there were only a limited number of large, powerful research computers in the country, and that many research investigators, who should have access to them, were geographically separated from them. Citation: Charles Herzfeld, ARPA Director (1965–1967) as quoted here: https://arstechnica.com/information-technology/2019/10/50-years-ago-today-the-internet-was-born-sort-of/ https://arstechnica.com/information-technology/2019/10/50-ye...

Animats 4y ago
MILNET actually was. It was 5-connected - each node had at least 5 connections to other nodes. Same technology as ARPANET, but DoD-only.
KineticLensman 4y ago
Don’t have refs to hand but IIRC the concept of packet switching was developed theoretically as a candidate for a military network but was deemed impractical by the Pentagon’s telecoms providers. However, the theory papers were picked up again by the early Arpaner designers
- mlyle 4y ago
  https://www.rand.org/pubs/research_memoranda/RM3420.html https://www.rand.org/pubs/research_memoranda/RM3420.html (1964) "We will soon be living in an era in which we cannot guarantee survivability of any single point. However, we can still design systems in which system destruction requires the enemy to pay the price of destroying n of n stations. If n is made sufficiently large, it can be shown that highly survivable system structures can be built—even in the thermonuclear era. "

Animats 4y ago

For instance, the idea that the Internet was designed to route around damage, and could also treat censorship as a form of damage, and therefore it was resistant to censorship. Yes. That comment was actually made about USENET, not the ARPAnet. Because USENET really did that. When two USENET nodes connected, they compared lists of message IDs, and each one requested any messages the other had, but they didn't. The result was distributed flooding - everybody eventually got everything in any group to which they subscribed. If someone censored the messages going across a link, that would prevent the message traveling over that link. But if any other possible path existed, two nodes would eventually compare messages lists, note the missing ID, and transmit the missing message. So censorship was only possible if you controlled all possible connections between two networks. This came up at Stanford in the 1980s, when some IT administrator (Ralph Gorin?) tried censoring rec.humor.funny. But there was one professor (McCarthy?) who had his own USENET node both on the Stanford network and connecting to the outside. It didn't have much capacity, but as long as there was even one low-bandwidth path, any censored messages would eventually take the slow path and arrive. So censorship was futile. But so was spam filtering.

Animats 4y ago

A packet-switched network is less necessarily centralized than a circuit-switched network. This is not actually true. The AT&T Long Lines network was originally very decentralized. When it was first fully automated, there was no central control point. There were 10 regional control centers, but they were not essential to most calls. Eventually there was central control, out of Bedminster, Pennsylvania. Originally all it did was collect data and send out new routing tables every few minutes. It didn't process any calls. Previously the routing tables had been fixed, changed maybe once a month. Calling pattern changes on holidays were a serious cause of overload. AT&T used to run ads asking people to make their Xmas calls earlier in the day. There's "A History of Science and Engineering in the Bell System"[1], written by the old-timers at the AT&T breakup to record how all the old stuff worked. A general understanding of how the more advanced electromechanical switching systems worked is useful in designing high-rel systems. Number 4 and Number 5 crossbar systems had system reliability much higher than component reliability. In the entire history of the Bell System, no electromechanical central office was ever out of service for more than 30 minutes, other than due to a natural disaster or major fire. That record has not been maintained in the computer era. [1] https://www.amazon.com/History-Engineering-Science-Bell-System/dp/0932764061 https://www.amazon.com/History-Engineering-Science-Bell-Syst...

dredmorbius 4y ago
A distinction though is that circut-switched loses connectivity when a circuit is broken. Packet-switched does not. You see some packet loss, there may be increased latency or reduced bandwidth, but the packets can still be routed without session loss so long as some valid routing exists. (For TCP --- I'd tell you the UDP joke but you might not get it ...). To that extent, the packet-switched network is more resilient, and this was a conscious design consideration, as can be seen in Paul Baran's original packet-switched networking documents: Simulation results are shown to indicate that highly efficient routing can be performed by local control without the necessity for any central—and therefore vulnerable—control point. The ten-part series starts here: https://www.rand.org/pubs/research_memoranda/RM3420.html https://www.rand.org/pubs/research_memoranda/RM3420.html The series as a whole is desribed and linked here: https://www.rand.org/about/history/baran.html https://www.rand.org/about/history/baran.html

kragen 4y ago

When John Gilmore said in the early 01990s, "The Net interprets censorship as damage and routes around it," he was talking about USENET, not the internet. USENET is a peer-to-peer gossip network, like Bitcoin. John Gilmore is a techno-utopian, a libertarian, and a technologist; he was Sun employee #5, he designed BOOTP (the basis of DHCP), and he wrote substantial parts of GCC and GDB. This saying of his was quoted in Time Magazine in 01993 https://web.archive.org/web/20210408023213/https://kirste.userpage.fu-berlin.de/outerspace/internet-article.html https://web.archive.org/web/20210408023213/https://kirste.us... which seems to be a bit confused about the relationship between the internet and USENET. I don't know how far back it dates, but I don't think he made it up for the article. The claim is, strictly speaking, true of gossip networks like USENET. If node A talks to nodes B and C, which also talk to each other and also to another node D, and node A utters a message that node B chooses not to distribute, the situation from D's perspective is precisely the same as if node B were temporarily down. Node C will still get the message (directly from A, rather than from B) and node D will still get it from C. You can find a lot of technologists who are also techno-utopians and libertarians making grandiose claims about the decentralizing influence of the internet in the cypherpunks archives from 01992 and 01993. At the time, the list was hosted on Hoptoad, John Gilmore's personal machine, in his house. I think there are a lot of ways that the internet architecture is less centralized than the PSTN that aren't related to packet-switching versus circuit-switching. Regular users aren't allowed to run traceroute on the PSTN (though some phone phreaks used to do that sort of thing, and MPLS means traceroute means less than it used to). The PSTN doesn't have a public whois database. It won't delegate even a class C of 256 phone numbers to your PBX for local use (though that was becoming difficult to get on the internet too by the mid-01990s), much less 2^48 phone numbers. You can't get a PSTN AS number so you can multihome your PBX to two different long-distance providers. Historically there was no mapping from logical names to phone numbers, so to call someone on the phone you had to dial an area code and exchange that identified their local phone company and its exchange, so user-visible addresses were inherently owned by local regulated monopoly providers, not by users themselves. Moreover, ideologically the deployment of new applications on the PSTN was centrally controlled and administered by the phone company, and they would run on the "switches" in the buildings without windows, while on the internet any random person could write and run an experimental server on their node, at least if their sysadmin was cool with it. Until the Carterfone decision of 01968 it was illegal in the US to plug even third-party telephones into the PSTN, and direct-connect modems didn't become legal until 01975. After that, lots of innovative applications of the PSTN were actually deployed in a decentralized way, such as answering machines, fax machines, and BBSes, but they were very limited in the features they could provide. By comparison, the early internet was wildly decentralized. Packet-switching does mean that a single computer can directly talk to an arbitrarily large number of other computers at once, instead of just however many circuits its connection supports. But a graph of degree three can already be a robustly cyclic connected graph, so three circuits per node would be enough. Of course, the phone company only provided most users with one at a time, with thousands of milliseconds of latency to establish a new circuit. So it wasn't so much that we imposed decentralization retroactively on the work of the 01970s and 01980s. It was more that we observed that it had resulted.

nl 4y ago

"Decentralization" has many meanings and often organizational and technical decentralization are conflated. The decentralization talked about in 1990's Wired was both because the technology wasn't controlled by a single company (unlike competitors like AOL or MSN).