On Thu, 22 Mar 2001, Gilbert Ramirez wrote:
> Guy Harris wrote:
> >
> >
> > 5) Put the existing dissector base into libethereal, which would also
> > include "register.c".
> >
>
> This is the right way to approach it, at first.
> Eventually it would be nice to organize the dissectors into
> large plugins, so that the "ipx" plugin has the dissectors for
> ipx, spx, ipxrip, ipxsap, nbipx, and ncp, while the "ipv4"
> plugin has icmp, ipv4, tcp, udp, or whatever.
>
> Moving the dissectors to an epan/dissectors directory
> would be appropriate, and would finally clean up our
> top-level directory. Who wants to do all that cvs work? :)
>
> --gilbert
>
Actually, I don't think that putting the dissectors into libethereal
is the right long term solution ( which is not to say that it isn't
the right short term solution). In my mind libethereal provides a
framework for doing useful work with dissectors. Dissectors should depend
on libethereal. Applications should depend on libethereal ( and
libwiretap, but it's pretty self contained already) If we put the
dissectors in libethereal we seriously risk
coming back to a place where libethereal has dependencies on dissectors,
which ( with the one dispensation for dissector_frame, which one way or
another can't be avoided ) I don't think is a very good idea.
I am more in favor over the long term of finding a way to compile the
dissectors either as individual plugins ( or possibly clusters of
plugins for related modules that make sense together ) or completely
linked into a static ethereal binary.
Moving towards a plugin architecture also forces us to remove compile
time dependencies of dissectors on each other. Which strikes me
as a good direction in general.
This being said, I don't think our current plugin architecture is quite
up to handling a lot of plugins with run time dependencies on each other
as we currently don't have a mechanism for resolving those dependencies
during plugin initialization.
Clearly this will be a lot of work. I'm willing to do a lot of it. I
just want to make sure calmer heads review the idea to make sure it's
the right thing to do for ethereal as a project.
Ed
