The Internet was starting to spread rapidly in the 1980s. Finland was buying CISCO routers like crazy. In the USA, TCP/IP was taking over as the preferred network protocol. But in Europe and Japan, there was a decision to define their own data network standard.
-
-
Show this thread
-
This was OSI, and it had at least one elegant theoretical idea, which was the protocol stack -- a logical layers of how you go from copper wires to network applications.pic.twitter.com/mxOJi2cqJ8
Show this thread -
I was at Bell Labs while OSI was happening, and I implemented some parts of the ASN.1 data-marshaling standard. This was sort of a binary version of XTML, an complicated way to define streams of data including text, integers, floating point numbers, and labels.
Show this thread -
I heard a lot of stories about OSI...mostly tales of disaster. A visiting researcher from IBM told us they built an X.400 email server, and they were appalled that, on a powerful mainframe computer, it took 1 full second of processing to send a single message.
Show this thread -
ASN.1 had optional formatting schemes...for examples you could implement strings as null-terminated and/or length/data format. X.400 servers built in Germany and France were unable to exchange messages, because they each close a different string format.
Show this thread -
In the internet, IP is the protocol for sending data packets, but without guarantee of delivery. The TCP protocol established a reliable data stream using IP...it put packets in order, tossed out duplicates and asked for re-transmission of missing packets.
Show this thread -
In OSI, the analogue to TCP was called TP4. Early implementations of that complex protocol were so slow, it took longer to send one packet than the timeout interval that TPC used to give up and close a connection.
Show this thread -
OSI was an expensive failure; not only in terms of the cost of the specification and development, but in terms of the opportunity cost of delaying the deployment of the internet in Europe and Japan. Trust in standards committee was sullied by the project.
Show this thread -
Fortunately, all of this happened well before 1995 when the internet became a mass media used by regular citizens. So I don't think the theory and experience learned from OSI were a disastrous waste.
Show this thread -
And while we are talking about fumbling the future, I should also say something about AT&T, where I worked. The Army once asked them to take over and manage the ARPANET, and the company declined.
Show this thread -
Even while I was there in the 1980s, there was hostility and derision of the idea of packet switched networking. The 1127 research lab (where UNIX was created) built their own circuit-switched computer network they called Datakit.
Show this thread -
The lab I was in (1138) was something of a pariah, because we used ethernet and ran TCP/IP and Berkeley UNIX on our computers. We also got scolded by the communications-workers union, because we installed the ethernet cable ourselves.
Show this thread -
AT&T's physical network was impressive technology: an all-photonic network with lithium-niobate crossbar switches, laser amplified undersea cables, and lossless token ring networks that serviced cities.
Show this thread -
I saw a presentation of this system that included making fun of TCP/IP and laughing at the idea that AT&T would support the internet. Ironically, the internet still runs on top of that photonic network infrastructure.
Show this thread
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.