What plans are there to extend the security model in Juju? We see two significant issues with the current model - no internal firewalling within an environment, and the inability to expose ports to systems outside the environment selectively.
Within an environment it appears that all machines can access any port on any other machine in that environment. For example, looking at the security groups in a Juju environment deployed on AWS, each instance launched by Juju is in a juju- security group with all UDP and TCP ports open to that security group. This means that if an attacker compromises any system in the environment they can probably go anywhere.
Also, the only way to expose ports outside of the environment is to expose them to any external system - there is no way, for example, to restrict access to a set of IP addresses.
Both EC2 and Openstack support internal firewalling and selectively opening ports, and other orchestrators such as Chef and Puppet support these features, so it would be good if Juju could as well.