5 ms·

The failure of Minitel lies mainly in the underlying networking model, and resulting capability to become a service provider. Minitel was based on x25. Contrar

The failure of Minitel lies mainly in the underlying networking model, and resulting capability to become a service provider. Minitel was based on x25. Contrary to IP, x25 was a network where the terminals are dumb and the network needs to be smart. This means that there was an incredibly (for the time) steep learning curve to transition from a Minitel user to a Minitel provider. This meant that the technology could not be easily studied and tested in a garage (even if some like me did), not impossible but tougher than with IP. You know the rest of the story. There are also more minor factors that lead to fall of Minitel but mainly it was IMHO a closed vs. open model root cause. I was heavily into both service and tech of Minitel at that time, AMA.

JPLeRouzic 6y ago
> x25 was a network where the terminals are dumb and the network needs to be smart. At the time that was a common statement, but I fail to see why X25 mandates dumb terminals. However a profound difference was that X25 used streams whereas TCP uses packets. A stream is akin to a TCP socket. It is a fundamental difference that stems from the fact that a X25 data exchange is safe from errors. https://farsite.com/X.25/X.25_info/X.25.htm https://farsite.com/X.25/X.25_info/X.25.htm
- philprx 6y ago
  It’s true that most applications were stream oriented, and Minitel was, as the users’ sessions were only bytes (in videotex encoding) rendered by the Minitel as a serial terminal integrating a modem (1200 bauds reception, 75 bauds emission!). There was a smart gateway node that translated complex x25 from the Minitel service providers nodes into simple bidirectional stream of bytes to the Minitels, it was a specialized type of packet asssembler disassembled (PAD) called a PAD Videotex (PAVI). Now x25 did have packets, Minitel terminals did not. So the PAVI and x25 servers were working very differently and with more complexity (hence smart) than Minitel serial terminals (hence dumb). It’s not that x25 mandated dumb terminals.
  - non-entity 6y ago
    Haha. I've been "working" on an X.25 implementation on a PC and it took me an embarrassingly long time to figure out what the hell a PAD was ever needed for.
    - icedchai 6y ago
      My understanding 30 years ago was a "PAD" was the TCP/IP equivalent of a "terminal server", translating serial streams to network connections. This may be an overly simplistic view, I don't know. It's been a very long time! Are there any public x.25 networks still operating?
      - non-entity 6y ago
        > Are there any public x.25 networks still operating? Not that I know of. My plan if to use old line simulators and other tools i can find on ebay and telcom equipment reseller sites to create a "fake" private one. This is one of those projects that I'm working on purely out of novelty interest :P.
        icedchai 6y ago
        Sounds cool! Perhaps you could create the x.25 equivalent of HECNet, the hobbyist DECnet.
- rjsw 6y ago
  OSI could run over X.25. As you say, it doesn't mandate dumb terminals. I worked for a couple of startups that ran services over X.25. One road not taken could have been to define a structured document format in ASN.1 that would be displayed by the terminal.
- eternalban 6y ago
  You are confusing connection-oriented phase of x.25 with “streams”. X.25 is a “packet-switched” “connection-oriented” protocol. [which p.s. makes sense for traditional telecom’s view of the world, which as you know is circuit-switched. One of the main beefs of telcos in regards to IP was traffic management. Compare X.25 congestion management with IP. ] Some kind soul has delved into the details of modalities of networking protocols: https://stackoverflow.com/a/43543284 https://stackoverflow.com/a/43543284
- C1sc0cat 6y ago
  I think its more correct to say x.25 was virtual circuit based - over a packet based system. x.25 had a MFT (Metric Fuck Ton) of error checking (at each layer). I used to work in OSI based networking - debugging OSI issues was an art - no wireshark back then However You could do some cool things with osi based networking binary attachments natively supported, years before the internet had it. Interworking between legacy system like Telex, I used to have my own telex number and answerback and could have had an x.400 email address at the country level. Ps id take site you linked with pinch of salt no mention of PAD's for example.
jillesvangurp 6y ago
The failure lies in the ownership model: a telco company trying to build a walled garden. Other walled gardens failed as well of course including some that were basically much more competitive in terms of networking. E.g. AOL failed a few years later because the world outside their walled garden was more interesting. It never really stood a chance of surviving despite being much more similar to the internet at that time. Operators trying to build walled gardens is a pattern that repeated itself until it became such a low margin business that the only operators able to make a profit were those merely offering full internet over 3G and 4G and not even bothering with the black hole of R&D needed to keep up with what was happening elsewhere. There were attempts in both standards to grandfather in a lot of stuff that is now exclusively done over IP but none of it ended up being popular. The history of mobile phone operators is just a long succession of attempts to copy things from the internet at the time and trying to turn them into walled garden things. Each of those things failed. We're not watching TV via operator provided services. We're not doing video calls via operator provided services, their game portals have long died and faded away. We're not browsing any WAP sites. Instagram wiped out what little adoption MMS ever had. None of it stood a chance. The incentives and ownership were just wrong and the role as operators as essential middlemen just never panned out in the real world as they hoped. Many telcos of course managed to convert themselves into successful mobile operators ultimately but their role was diminished to providing network access. My android phone has a phone stack. The only stuff that comes in via my phone number is basically recruiter cold calls. It's a small internet device with some quaint legacy features that I rarely use. Really, numeric identifiers that you have to punch in? In 2020? How is that still a thing? I remember memorizing shit like that a few decades ago. So, minitel is merely an early and probably expensive incarnation of that. France telecom subsidized the devices and had to earn it back via regular network charges. I doubt that generated anywhere near what they hoped in revenue.
- philprx 6y ago
  Absolutely all true, these notions actually share the same model: centrally controlled, no control at terminal / no open access for new entrants, consumption devices are locked to consume only, hard to “become center” or provider, etc. The networking model was a logical consequence (and practical implementation) of the the centrally controlled walled garden you describe. Many hackers were trying to escape it by becoming their own providers and also connecting with other international x25 servers and foreign hackers. For example, Assange was actually from these times. Nevertheless, it did generate much more revenue than what they initially envisioned.
anthk 6y ago
TCP/IP can work over AX25, hamradio users know it well.
- dfox 6y ago
  AX.25 is to a large extent completely different beast than what people mean when they talk about "X.25 network". X.25 as a specification is to a large extent an "API" and not particular network implementation. AX.25 is complete protocol stack that implements network with superficially similar and heavily simplified API.
rbanffy 6y ago
> The failure of Minitel lies mainly in the underlying networking model, and resulting capability to become a service provider. Not sure it's that simple. While you couldn't easily host your own services on the telco mainframes, there was nothing stopping companies from offering services over the phone. We had a Minitel service where I grew up in Brazil and numerous BBSs supported the 1200/75 dial in speeds and had specific UIs (or, at least, "ASCII" art) for the terminals. Supporting the cheap and easily available modems was a no-brainer for the BBS owners, even though they needed to do bit-banging to make the asymmetric speed work. Think of the terminal as a primitive but ubiquitous browser and the phone system as a circuit switched network and the dial in services as websites and you have the 80's web.
conradfr 6y ago
I'm not sure it can be described as a failure, it made a lot of money and provided many jobs in France (I think my father worked on it during the 80s as a telco engineer). The last time I had to use it must have been in 1999 as the only choice to enter your college preferences. And there was "Minitel over the internet" for some time. Apparently it only closed in 2012! Transitioning the country to the Internet was botched in a typical French fashion though.

JPLeRouzic 6y ago

> x25 was a network where the terminals are dumb and the network needs to be smart. At the time that was a common statement, but I fail to see why X25 mandates dumb terminals. However a profound difference was that X25 used streams whereas TCP uses packets. A stream is akin to a TCP socket. It is a fundamental difference that stems from the fact that a X25 data exchange is safe from errors. https://farsite.com/X.25/X.25_info/X.25.htm https://farsite.com/X.25/X.25_info/X.25.htm

philprx 6y ago
It’s true that most applications were stream oriented, and Minitel was, as the users’ sessions were only bytes (in videotex encoding) rendered by the Minitel as a serial terminal integrating a modem (1200 bauds reception, 75 bauds emission!). There was a smart gateway node that translated complex x25 from the Minitel service providers nodes into simple bidirectional stream of bytes to the Minitels, it was a specialized type of packet asssembler disassembled (PAD) called a PAD Videotex (PAVI). Now x25 did have packets, Minitel terminals did not. So the PAVI and x25 servers were working very differently and with more complexity (hence smart) than Minitel serial terminals (hence dumb). It’s not that x25 mandated dumb terminals.
- non-entity 6y ago
  Haha. I've been "working" on an X.25 implementation on a PC and it took me an embarrassingly long time to figure out what the hell a PAD was ever needed for.
  - icedchai 6y ago
    My understanding 30 years ago was a "PAD" was the TCP/IP equivalent of a "terminal server", translating serial streams to network connections. This may be an overly simplistic view, I don't know. It's been a very long time! Are there any public x.25 networks still operating?
    - non-entity 6y ago
      > Are there any public x.25 networks still operating? Not that I know of. My plan if to use old line simulators and other tools i can find on ebay and telcom equipment reseller sites to create a "fake" private one. This is one of those projects that I'm working on purely out of novelty interest :P.
      - icedchai 6y ago
        Sounds cool! Perhaps you could create the x.25 equivalent of HECNet, the hobbyist DECnet.
rjsw 6y ago
OSI could run over X.25. As you say, it doesn't mandate dumb terminals. I worked for a couple of startups that ran services over X.25. One road not taken could have been to define a structured document format in ASN.1 that would be displayed by the terminal.
eternalban 6y ago
You are confusing connection-oriented phase of x.25 with “streams”. X.25 is a “packet-switched” “connection-oriented” protocol. [which p.s. makes sense for traditional telecom’s view of the world, which as you know is circuit-switched. One of the main beefs of telcos in regards to IP was traffic management. Compare X.25 congestion management with IP. ] Some kind soul has delved into the details of modalities of networking protocols: https://stackoverflow.com/a/43543284 https://stackoverflow.com/a/43543284
C1sc0cat 6y ago
I think its more correct to say x.25 was virtual circuit based - over a packet based system. x.25 had a MFT (Metric Fuck Ton) of error checking (at each layer). I used to work in OSI based networking - debugging OSI issues was an art - no wireshark back then However You could do some cool things with osi based networking binary attachments natively supported, years before the internet had it. Interworking between legacy system like Telex, I used to have my own telex number and answerback and could have had an x.400 email address at the country level. Ps id take site you linked with pinch of salt no mention of PAD's for example.

jillesvangurp 6y ago

The failure lies in the ownership model: a telco company trying to build a walled garden. Other walled gardens failed as well of course including some that were basically much more competitive in terms of networking. E.g. AOL failed a few years later because the world outside their walled garden was more interesting. It never really stood a chance of surviving despite being much more similar to the internet at that time. Operators trying to build walled gardens is a pattern that repeated itself until it became such a low margin business that the only operators able to make a profit were those merely offering full internet over 3G and 4G and not even bothering with the black hole of R&D needed to keep up with what was happening elsewhere. There were attempts in both standards to grandfather in a lot of stuff that is now exclusively done over IP but none of it ended up being popular. The history of mobile phone operators is just a long succession of attempts to copy things from the internet at the time and trying to turn them into walled garden things. Each of those things failed. We're not watching TV via operator provided services. We're not doing video calls via operator provided services, their game portals have long died and faded away. We're not browsing any WAP sites. Instagram wiped out what little adoption MMS ever had. None of it stood a chance. The incentives and ownership were just wrong and the role as operators as essential middlemen just never panned out in the real world as they hoped. Many telcos of course managed to convert themselves into successful mobile operators ultimately but their role was diminished to providing network access. My android phone has a phone stack. The only stuff that comes in via my phone number is basically recruiter cold calls. It's a small internet device with some quaint legacy features that I rarely use. Really, numeric identifiers that you have to punch in? In 2020? How is that still a thing? I remember memorizing shit like that a few decades ago. So, minitel is merely an early and probably expensive incarnation of that. France telecom subsidized the devices and had to earn it back via regular network charges. I doubt that generated anywhere near what they hoped in revenue.

philprx 6y ago
Absolutely all true, these notions actually share the same model: centrally controlled, no control at terminal / no open access for new entrants, consumption devices are locked to consume only, hard to “become center” or provider, etc. The networking model was a logical consequence (and practical implementation) of the the centrally controlled walled garden you describe. Many hackers were trying to escape it by becoming their own providers and also connecting with other international x25 servers and foreign hackers. For example, Assange was actually from these times. Nevertheless, it did generate much more revenue than what they initially envisioned.

anthk 6y ago

TCP/IP can work over AX25, hamradio users know it well.

dfox 6y ago
AX.25 is to a large extent completely different beast than what people mean when they talk about "X.25 network". X.25 as a specification is to a large extent an "API" and not particular network implementation. AX.25 is complete protocol stack that implements network with superficially similar and heavily simplified API.

rbanffy 6y ago

> The failure of Minitel lies mainly in the underlying networking model, and resulting capability to become a service provider. Not sure it's that simple. While you couldn't easily host your own services on the telco mainframes, there was nothing stopping companies from offering services over the phone. We had a Minitel service where I grew up in Brazil and numerous BBSs supported the 1200/75 dial in speeds and had specific UIs (or, at least, "ASCII" art) for the terminals. Supporting the cheap and easily available modems was a no-brainer for the BBS owners, even though they needed to do bit-banging to make the asymmetric speed work. Think of the terminal as a primitive but ubiquitous browser and the phone system as a circuit switched network and the dial in services as websites and you have the 80's web.

conradfr 6y ago

I'm not sure it can be described as a failure, it made a lot of money and provided many jobs in France (I think my father worked on it during the 80s as a telco engineer). The last time I had to use it must have been in 1999 as the only choice to enter your college preferences. And there was "Minitel over the internet" for some time. Apparently it only closed in 2012! Transitioning the country to the Internet was botched in a typical French fashion though.