Some notes regarding the implementation of the SCAMP protocol

1: The mechanism for handling infinite subscription forwarding in small
networks is done differently, I use a TTL which allows a subscription to be
forwarded at most 2*N times (N is network size). This is a pure rule of thumb,
but in the simultions it fullfills the necessary role.

2: The heartbeat mechanism is not implemented, as we agreed.

3: Failure handling is absent, that is, communications are one way,
without feedback and messages sent to failed nodes are simply lost.

4: The unsubscription implementation sticks to the one described in the paper.
It has known flaws, that is, it erodes the consistency of inViews, which
results in the erosion of the whole graph on the long run. Some of these
effects can be fixed by more communication as I described earlier, I haven't
implemented it waiting for a consensus.

5: The simulation of lease is still an open question. I adopted the
version in which each node removes expired subscriptions from their own views,
the other version, in which nodes remove their own expired subscritions from
the views of other nodes is commented out.

