Hi,
On Wed, 19 Apr 2000, Dmitry Jemerov wrote:
> Actually, I'm not sure that the FreeSCI code should be split into many
> libraries. The only thing that makes sense to me is creating a separate
> library for code that will be used by external console-mode tools (like
> scidisasm or possibly some SCI development tools).
>
> Other than that, a single dynamic-link library containing all the code
> except for the uppermost level (main.c and config.c) should work just fine.
Considering that we have several separate helper programs (none of which
is really performance critical), using a dynamically linking library makes
some sense. It has three disadvantages, though:
- People who have to run it from their UN*X $HOME will need to tweak their
LD_LIBRARY_PATH
(Alternatives like using ldopen() to link manually during run-time are
possible, but not very portable).
- Statical linking is available on all platforms, while dynamic linking
might not be (IIRC, DOS needs a helper program to do this, and some
other semi-obsolete OSses don't like dynamic linking either (IIRC,
again)).
- End users aren't interested in the helpers, they just want to play their
games. They only need the sciv executable anyway, so a libsci.so would be
overkill for them.
None of these apply for Win9x or NT (except for #3, which isn't really
that fatal), so dlls make sense there. I'm not convinced that it's the
optimal solution in the generic case, though.
llap,
Christoph
