What I was visualizing was basically turning all dissectors into
plugins, too, since they would be dynamically loaded on-demand. But we
need to know when to load them, so we have to load *something* from
each dissector. That's what I meant about the registration stuff.
Split the main code of the dissector from the registration code.
My motivation is not the same... I'm more worried about the memory
being used by wireshark, rather than the startup-time. Why load some
of the bigger dissectors into memory if I'm never going to use them?
thanks,
--gilbert
