The network effect is the notion that economies of scale grow exponentially as the number of nodes increases. The classic example is the telephone. One telephone on the planet is worthless; a hundred or a thousand, much better. Nowadays, area codes breed like rabbits, and people talk to themselves on the streets in a manner that would have caused them to be locked up only a few years ago. Perhaps less benign is the use of lowest-common-denominator tools "because everyone else uses them," even if the tools themselves don't inspire excitement. Four-letter model-building and word-processing tools spring to mind as examples.
Standards, with interchange and interoperability of course, mean that you can buy each of these tools from a different vendor, thus bypassing potentially monopolistic vendors who might otherwise become slothful. This allows you to pick the best tools for the job. The existing UML, QVT, and other MDA-related standards create the network effect because the existence of multiple vendors of standards-conforming tools reduces vendor "lock-in," the unhappy (for the buyer, that is) situation that you can only buy tools from a single vendor.
More standards will be required as MDA matures. These will address many technically detailed topics, such as standardization of metamodels; the use of UML profiles versus true meta-case tools that allow modelers to define their own graphical notations for their metamodels; and the standardization of mapping functions, platforms, and the architectures to which they map. There will be a need for the perfect integration of tools across all levels: an architecture construction kit, architecture wizards that help us find and experiment with different architectures. "What kind of persistence mechanism would you like to use a relational DB, OODB, file, hierarchical, host-based? What component framework would you like on top J2EE, .NET, none? What type of front end would you like to build Web-based, 3270, fat client?" Damn! There's that animated paper clip again!
Model-driven architecture, of course, is the name of the game. Not only must there be standards for the UML, the MOF, QVT, and all of the rest, but it's also critically important that tools built to these standards also fit together within that architecture and so create a complete model-driven development environment. This set of tools, loosely sequenced, constitutes a tool chain. MDA intends to build the architecture for this tool chain.