> At the start of the GOSIP nonsense, that might have been a reasonable > charge. By the middle, there were at least as many ISO OSI > applications as there are now IPv6 applications, and there was a > lot of real OSI traffic in Europe. (A "lot" for that era if not > today.) Major host vendors were supporting OSI protocols by the > end of GOSIP as well as they are now supporting IPv6. OSI support > was a required checklist item in a lot of sales situations. Well, you have to seriously qualify what you call "a lot". The only non-TCP-IP network service that got widely deployed in Europe was X.25. Most of its use was not strictly OSI: asynchronous modem connections according to X.28 and X.29, or enterprise networks using X.25 as a lease line replacement, much like modern-day VPN. Some of that is still in use today, e.g. for the French "Minitel" service. The only OSI service that had significant deployment was the mail system, X.400. It was used for some research networks, very often through gateways with SMTP mail. It was also used for exchange of documents between enterprises in specialized EDI networks. An aggravating factor in the GOSIP story was the lack of agreement on the network + transport stack. The US proposed the IP-like CLNP and the TCP-like TP4, the European X.400 deployments mostly used X.25 and TP0 (null transport), and ECMA was pushing a null network layer combined with TP4 for European enterprise networks. The academics who were using X.400 and experimenting with X.500 quickly moved to a combination of TCP-IP and TP0 (RFC 987, RFC 1006), because TCP-IP was more accessible than either X.25 or CLNP. Most X.400 MTA used in academic networks where dual stack, i.e. able to translate between X.400 and RFC 822, and most of the traffic moved quickly to plain SMTP. Similarly, the X.500 experiments resulted in LDAP, i.e. a TCP-IP application. Bottom line, there never was a significant usage of CLNP in Europe. -- Christian Huitema