As pointed out earlier, scientists want to collaborate at different levels. They may want to share their raw data, in which case it may be more efficient to replicate the data set at the remote sites for faster processing. However, they may also opt to allow indirect sharing of their data through the visualization results. That is, don't share the raw data, just the visualization results of the data. In this case, the geometric primitives that represent the visualization results need to be distributed to each member of the session. This also implies that other participants can grab the spray can to visualize other regions of the same data set. Recall that spray requests are sent to the host which created the spray can (i.e. where the data resides) to create and send out the visualization results. Yet another form of sharing is to provide images of the visualization but without granting access to manipulate and spray the cans.
Different levels of data sharing is closely related to service matching. That is, participants in a session may have workstations of varying degrees of power. The goal of service matching is to provide the fastest feedback possible to users in a session within the constraints of the machines, the network, and the levels of data sharing the users set for each data set. It also implies that the session should avoid being bogged down by a participant on a slow machine. If one of the participants was on a slower machine (e.g. an X terminal with no graphics accelerator), the actual visualization work could be done remotely and then compressed image data sent to the slower machine. This is practical to the extent that the participant on the slower machine is willing to share raw data. Otherwise, requests to spray that participant's cans would become a bottleneck in the session.