Gilbert Ramirez wrote:
> One could argue that since you have no, to pick on one dissector, token-ring
> on your network, for example, that we should make token-ring either a plugin or
> a configure-time option. You might want SMB or NFS to be plugins too, since
> they are large dissectors. We haven't discussed that possibility at all.
I'd really love to see the capability of the code being written such
that at compile time, one could choose between a module being directly
linked into the code and being a module. I would say - use the same
plugin mechanism for calling the dissectors, but have them linked in
directly instead of being dynamically loaded. This has the added benefit
of supporting static linking in a reasonable fashion.
Also related to this question - it would be nice to see a global
"enable/disable protocol" option for every protocol. This should be easy
to manage generating the list (just use the list of protocol keywords
that have been registered), but actually supporting disabling/enabling
the encoding would require that code modifications be made to check to
see if the protocol is enabled. Not sure if this would be best to be
done where the dissectors are called, or at the beginning of each
dissector.
Perhaps we could even have a default state for the enable/disable flag,
for example so we can include possibly buggy dissectors in the mainline
source without having them turned on, that way, people could test them
easily to report bugs, but if they had problems, they could just shut
the dissector for that protocol off temporarily.
-- Nathan
------------------------------------------------------------
Nathan Neulinger EMail: nneul@xxxxxxx
University of Missouri - Rolla Phone: (573) 341-4841
CIS - Systems Programming Fax: (573) 341-4216