[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: CSP and Petri nets?


>   What are the relative strengths of and relations between CSP and
> Petri nets?

I've always steered well clear of Petri nets.  In their strict form,
they give a data-driven parallelism (nodes fire only when input tokens
arrive from *all* sources) that is always deterministic.  To model
an ALT (say) that's listening and reacting to just one of many input
sources, you have to arrange dummy tokens to be continually moving.

I've always found them bewildering - especially when compared with
occam-style network diagrams.  Petri net nodes are simply modelled
by an occam process, but not vice-versa.  My feeling (I've never ever
done it!) is that you have to go to a lot of trouble to build a Petri
net model of an event-reacting CSP design ... and I don't see the
end benefit from all that trouble.  Petri nets have well-defined
mathematical properties, but so does CSP ... I don't know how Petri
net analysis tools compare with those from CSP nor whether they can
deliver as many useful properties ... need input from proper Petri
experts ...