As I said, all APIs are technically protocols, although the IETF has an extra implied connotation of distributed function. Now then...
on 8/15/2003 11:55 AM Larry Osterman wrote: > For that matter, one could argue that IMAP isn't a protocol, since it's > just an expression of the c-client API set. IMAP is described as a set of distributed functions. Although it could be implemented as local API, the specs go beyond that. The discussion on TCP window size management is an obvious (but unnecessary) example. > For example, how do you handle NetBIOS over TCP (RFC 1001/1002)? It is > a set of protocols built on top of TCP, but exactly express the NetBIOS > API set semantics. Is NBT a protocol by your definition? Without refreshing myself on all of the components, parts of it are. The sum of the parts also are. Things like TDI would not be however. ODBC is probably a better parallel. Some of the drivers are protocols, but the API to the drivers is not. > If the only thing that makes NBT a protocol the fact that there's an > protocol specification for it, I can assure you that there's a protocol > specification for the MAPI RPC protocol - it's not open, but it exists. The "API-extension protocol" is something else. -- Eric A. Hall http://www.ehsco.com/ Internet Core Protocols http://www.oreilly.com/catalog/coreprot/
