siiky
2024/05/13
2024/05/17
2024/05/17
whitepaper,petri_nets,distributed
A basic implication is that if a place p resides at one location and if there is an arc from p to t, then t is necessarily required to reside at the same location as p. Tokens can thus always be seized and consumed in zero time. As a consequence, arcs from places p to transitions t (such as in Figure 3) ought to stay direct arcs and in any implementation, they are disallowed to be expanded as was done in the middle of the figure. By contrast, an arc from a transition t to a place p, as shown in Figure 4, does not prevent t and p to reside at different locations, since producing tokens in non-zero time does not compromise choices. As a consequence, such an arc is allowed to be prolonged indefinitely.
p4 checking whether a transition t is enabled must happen on the node that owns t's pre-set. It implies that a transition t cannot consume from places of different nodes. Additionally, while in theory transitions produce tokens instantaneous and simultaneously, it is not necessarily true in practice where a transition may produce to places of different nodes. However, it isn't problematic semantically.
Can a system described by some Petri net exhibiting concurrency actually be distributed?
p17