A proxy is a kind of event buffer that allows to pass some kind of barrier. The barrier can be a thread or a network for example. A proxy completely decouple the source from the sink. The proxy buffers the source events and when possible it sends them to the listeners at the sink. In other words, a proxy is indirect, non synchronized and non blocking.
The usual source/sink mechanism is synchronized, direct and blocking : when the event occurs, all listeners are called, and we have to wait they finish to process these events to continue and send new events.
With proxies, there is a buffer often compared to a mail box. Each event produced as source is buffered and when the sink is free to receive these events it can check the mail box and empty it, thus receiving the pending events. This way of doing is completely non synchronized and non blocking (due to the mail box).
This way of doing allows for example to pass a thread frontier with a minimum of synchronization : only the mail box has to be synchronized. And the source and sink can most of the time run in parallel. Without such a proxy, we would have to synchronize the whole graph, and threads would consume their time waiting one another since most of the work in GraphStream is centered on graphs.
For networks, this is the same thing : events are buffered before sending them to the network. When the other end is ready it can check these events in one operation.
However proxies have a limitation : they force the receiving end to check for events regularly. This can be compared to "pumping" since the whole GraphStream metaphor is a set of sources, pipes and sinks. Here instead of flowing freely, the event stream must be pumped manually to receive it. This is however most of the time not a problem since most work on graphs in GraphStream is dynamic and runs iteratively.
| |
| |