A common doorway for incompatible computers
In 1969, the computers joining ARPANET were not built to speak the same internal language. Each mainframe had its own hardware and operating system, so connecting every machine directly to every other would have required custom links and translation rules for each pair.
The network’s answer was to put an Interface Message Processor, or IMP, between each host computer and the communications lines. A host did not need to imitate another institution’s mainframe. It needed to exchange messages with its local IMP according to a shared interface called 1822.
What the IMP did with a message
The 1822 specification described how a host handed a message to an IMP and how the IMP handled that handoff. The IMP then forwarded the message across the network toward the IMP attached to its destination host. Along the way, IMPs could pass traffic from one to another rather than requiring a direct physical connection between every pair of mainframes.
This arrangement separated two jobs: each institution’s computer could keep its own design, while the IMPs managed delivery across the shared network. That made the IMPs practical intermediaries—not translators that made all computers alike, but gateways that gave them a common way to send and receive network messages.
A design that made expansion manageable
Without that boundary, adding a new kind of computer could mean building and maintaining special connections to existing systems. With an IMP-based network, a new host could join by meeting the local interface requirements. The network could therefore grow without forcing its participants to replace their mainframes or surrender their proprietary systems.
The important trade was standardization at the edge: the computers remained different, but their connection to ARPANET followed a common procedure. That small layer of agreement let otherwise incompatible machines take part in one network.
