On Mon, 9 Jul 2001, Guy Harris wrote:
> > That being said... just because it's being done other places doesn't mean
> > it's a good idea. Wouldn't it make more sense to build an api for
> > querying the preference values rather than querying them via
> > compile time linking?
>
> It might.
>
> However:
>
> > I'd really like to see as much of the compile time dependencies between
> > dissectors go away as possible...
>
> ...there are cases where it's not obvious that the dependency isn't
> somewhat intrinsic.
Absolutely agreed. There are time when the dependence is intrisic and
there is no sensible way of getting around it.
>
> For example:
>
> > > packet-rpc.h:extern const value_string rpc_auth_flavor[];
>
> The MOUNT protocol uses the same RPC authentication flavor values as
> does the ONC RPC protocol itself, and it'd clearly be bogus to have two
> separate value_string tables with identical contents; is it really worth
> it to add Yet Another API for that? It makes no sense whatsoever to
> build Ethereal with the MOUNT protocol dissector but not the ONC RPC
> dissector (given that MOUNT runs atop ONC RPC), so the MOUNT dissector
> depends, in effect, on the ONC RPC dissector no matter what you do.
>
> Similarly:
>
> > > packet-wap.h:extern const value_string vals_character_sets[];
>
> both WTLS and WSP presumably use the same character set codes here, and
> it's not clear that it makes sense to have some of the WAP family
> dissectors without the others.
>