3 ms·

The Internet already got out so fast it beat the Big Government Solution, which was OSI, so comprehensively that people these days don't even know that there wa

by msla 4y ago

The Internet already got out so fast it beat the Big Government Solution, which was OSI, so comprehensively that people these days don't even know that there was an OSI protocol suite, let alone that it got beaten by the TCP/IP stack. Back in the early days, the ARPANET stuff was seen as a sideshow, whereas OSI was going to be the Serious Network for Serious Users, with X.500 directory services supporting X.400 email and nobody would remember toys like LDAP and SMTP and POP. Of course, these days the idea that TCP/IP somehow implements a purely conceptual OSI is so ingrained that people get angry when you tell them the accurate history. So it goes.

petre 4y ago
The OSI model is too complicated. I learned about it in the CCNA course. That's why it continues to exist as a mere model.
trasz 4y ago
I believe OSI was at least as much corporate as it was governmental. And, like most corporate solutions it ended like it ended, with design by committee. IP on the other hand was neither, like much of successful technology. As a side note, the OSI layers are useful as a mental model, but trying to maintain that separation in actual implementation is a poor idea; that’s one of the things IP got right: it doesn’t really match OSI layering, and does cut through layers when it makes sense.
- msla 4y ago
  > that’s one of the things IP got right: it doesn’t really match OSI layering, and does cut through layers when it makes sense. It doesn't implement the OSI layering model, in other words, even if you can almost make that model fit if you ignore the top couple layers and handwave vigorously the rest of the time. If you look, you can find a four-layer or five-layer TCP/IP model, but RFC 3439 has a section titled "Layering Considered Harmful" and it's clear that the IETF doesn't always think in terms of layers. https://datatracker.ietf.org/doc/html/rfc3439 https://datatracker.ietf.org/doc/html/rfc3439 Of course, none of that stops the Seven-Layer Model (blessed be its name) from being perpetuated as if it were an absolute.

by msla 4y ago

petre 4y ago

The OSI model is too complicated. I learned about it in the CCNA course. That's why it continues to exist as a mere model.

trasz 4y ago

I believe OSI was at least as much corporate as it was governmental. And, like most corporate solutions it ended like it ended, with design by committee. IP on the other hand was neither, like much of successful technology. As a side note, the OSI layers are useful as a mental model, but trying to maintain that separation in actual implementation is a poor idea; that’s one of the things IP got right: it doesn’t really match OSI layering, and does cut through layers when it makes sense.

msla 4y ago
> that’s one of the things IP got right: it doesn’t really match OSI layering, and does cut through layers when it makes sense. It doesn't implement the OSI layering model, in other words, even if you can almost make that model fit if you ignore the top couple layers and handwave vigorously the rest of the time. If you look, you can find a four-layer or five-layer TCP/IP model, but RFC 3439 has a section titled "Layering Considered Harmful" and it's clear that the IETF doesn't always think in terms of layers. https://datatracker.ietf.org/doc/html/rfc3439 https://datatracker.ietf.org/doc/html/rfc3439 Of course, none of that stops the Seven-Layer Model (blessed be its name) from being perpetuated as if it were an absolute.