Re: [Adonthell-devel] Thoughts about modules

[Kai Sterker]

Am Sonntag, 15.02.04 um 20:14 Uhr schrieb Alexandre Courbot:

> I've begun implementing the current map model prototype in C++, under 
> the
> "landmap" module. Doing so, I've noticed that "landmap" will never 
> have to
> use something else than "base" and "event" (as could be excepted, 
> since the
> game mechanisms and user interfaces are now completely separated). 
> Worse, it
> appears that all our modules could actually be divided into two types 
> of
> modules: the game modules and the input/output modules.
[...]
> The interesting thing to notice is that the I/O modules are the only 
> ones to
> require a backend (like SDL) and the special starting method using 
> backend
> loading. What it means is that a server-only binary will never need to 
> load a
> single backend

Doesn't a server use network (which might depend on backends for 
portability)?


> Maybe we have to redefine our design here. I'd like to see something 
> more
> coherent, although I fail to imagine how it would look like. Any idea 
> about
> this?

If a server app really would have no need for a backend, then a 
slightly different design would be good. However, I am not quite sure 
what can be done. Either we'd add some #ifdef SERVER directives to 
libmain to exclude the backend code, or we could write a libservermain. 
But both alternatives aren't really that great.

OTOH, libmain was written to take care of initializing the I/O part 
where needed. If a server app doesn't need this, perhaps it best to not 
link it to libmain.

(Besides, if libmain would really include something like a game 
launcher and settings dialog, it would (a) require libgui and libinput 
and (b) should not appear in a server app.)


> PS: Jol and I will be present at FOSDEM 2004, if some of you go there 
> too, we
> could have a drink! ;)

Sorry, but I'm too busy ... :(

Kai