Telescript Security

Among the biggest concerns about Telescript is that an agent isn't much different from a virus. General Magic and AT&T have addressed this problem in numerous ways. The first is that Telescript is an interpreted language. A virus is typically executable code that inserts itself into a target machine; requiring agents to be interpreted at a destination precludes executing something unrecognizable or illegal.

As described above, Telescript agents are encrypted and have to pass an authentication barrier. Once past it, they must have a valid address to go to (for additional security, users can layer end-to-end encryption on top of that built into the network). A place can also refuse to accept an agent if it's asking for something the place doesn't want to provide.

Telescript supports a scheme, known as permissions, to regulate the activities an d life span of agents. It includes limits such as how long an agent can live, how many CPU cycles it can consume, and whether or not it can spawn children or clone itself. Permissions are negotiated among the agent, place, and PersonaLink network itself: The agent asks for a set of permits, and the place and the network publish the sets they are willing to grant. The minimum coincidence of the three becomes the set of allowed permits. Thus, a mailbox could grant an agent from its owner the right to view its contents and delete messages but would refuse those rights to an agent representing somebody else.