From a farther remove, the domain is merely something that obeys programs written in some instruction set and invokes capabilities that it has in due course come to hold. This is the remove described in the OSR architecture paper. This illusion is well neigh perfect and only the space bank is in a position to sense the truth. The space bank has no interest in revealing this situation and thus the secret is safe.
The factory very nearly hides the domain abstraction and would do so by a mere rewriting of the manual.
If there were legacy Keykos programs to run, written for another processor’s instruction set, it might be strategic to build domains that would execute the legacy instruction set. This construct could be implemented outside the kernel just as normal Keykos objects are. The interpreter would keep the processor state in ordinary pages, just as do ordinary Keykos objects. There would be no special arrangements with the real kernel regarding the new construct. On the other hand if there were hardware in the system to execute the old instruction set, then the kernel could support legacy domains directly.
There would be no domain tool for the new kind of domain. There would a domain creator creator for the synthetic domains, SDCC. SDCC would create creators of synthetic domains, SDC. An SDC would create a synthetic domain SD. SDCC, SDC and SD are polymorphic with the respective real domain counterparts except that details concerning process state orders on a synthetic domain would be appropriately changed.
Synthetic domain creators would recognize their yield just like real domain creators. The distinction between real and synthetic domains would be hidden to all except the implementors of the new domain type.