Ten years ago, conventional wisdom said that the computing industry would be hurtling toward wholesale adoption of the OSI standards by now. TCP/IP protocols were positioned as a stepping-stone to OSI. Likewise, SNMP was merely a precursor to the OSI network-management standards that were expected to gain world dominance.
But if you stop and take a look around, it hasn’t happened. In fact, it hasn’t even started to happen. TCP/IP continues to gain market momentum, and products based on OSI (Open Systems Interconnect) standards are few and far between.
In fact, rumblings can be heard within the industry that OSI as a whole is dead, and that TCP/IP, rather than acting as a stepping-stone to standardization, will itself become the standard of choice. Indeed, many users and vendors no longer talk about compliance with the OSI standards, but rather discuss technology implementations such as the Open Software Foundation’s Distributed Computing Environment (DCE) and Distributed Management Environment (DME).
That, of course, begs the question: Why hasn’t OSI fulfilled its promise? That’s a big question, and it requires a long and involved answer. Certainly, the fact that the OSI standards were largely defined for and by telecommunications and large system vendors — long before the advent of PC networking — explains a lot. But a more fundamental cause can be found in the way standards are created.
Before a technology can even be put on the TCP/IP track by becoming a Proposed Standard, for example, a working prototype of the technology must be demonstrated. For a Proposed Standard to become a Draft Standard, fully usable implementations of that technology are usually available for people to test.
Before a Draft Standard becomes an Internet Standard, usable implementations of the technology must be demonstrated, and the Internet community must wholeheartedly accept it. This process gives voice to the people who actually use the technology.
OSI standards proposals, on the other hand, are conceived as Working Documents. Working Documents weave their way through a bureaucracy. Once agreed upon, Working Documents become Draft International Standards. After more meetings and discussions, the document is promoted to International Standard. At no time during this process are the defining committees required to create prototypes or working implementations of the standards for evaluation. Product testing is left to the end, after the standards are de fined.
The overwhelming success of SNMP (Simple Network Management Protocol) –and the near total failure of CMIP — can be attributed to the obvious differences in the two processes of standardization. SNMP was proposed in 1986. The first implementations of the technology were demonstrated in 1988, and usable implementations of SNMP have been available since 1989.
The OSI management standards remain cumbersome, hard-to-read documents. The Network Management Forum (NMF), which was founded to guide implementation of OSI network-management standards, released its first specification in 1989 after several years’ work. Products based on the NMF specification have yet to be shipped.
Users desperate for network management have bought products based on SNMP, and SNMP continues to gain enormous momentum, while the OSI documents gather dust. Note that this outcome had little to do with technology and much to do with pragmatic approaches to real problems.
To be effective, standards must be defined in terms of the reality of everyday use, and they must accommodate the installed base of systems. OSI violates both maxims.
Now, before you condemn me for heresy, let me say that the OSI standards will influence product development. But those products won’t be pure OSI; they’ll be environment-specific products based on standards. And they’ll achieve success only if they offer a pragmatic approach to user needs.