[MUD-Dev] Re: PDMud (was Re: Bruce Sterling on Virtual Community goals)

[Jo Dillon <emily#thelonious,new.ox.ac.uk>]

James Wilson (jwilson#rochester,rr.com) spake thusly:
>
> Dunno if this helps - several years ago I implemented a bare-bones server
> using dynamically-loaded native-code modules. Each module had a set of
> dependencies which would be automatically loaded before it. The dynamic-linking
> phase connected symbols in each module to symbols in other modules, and
> everything was happy. (Note that some systems' dlopen only resolves symbols
> from dynamically-loaded modules to the main executable, so inter-module
> linkage had to be done by hand with dlsym. I think the issue was with the SGI
> dl* functions.) There was no need for message-based communication;  you could
> directly call methods and functions in other modules. The bootstrap module
> loaded other modules by name, so didn't have to know anything about them (this
> was controlled by a text file).

  This sounds like a sensible way to do things. I really do think that
we should pay attention to speed as far as possible; we should be able
to cope with computationally-intensive muds which implies as much efficiency
in the server engine as possible.
 
> This worked pretty well, but forced me to write plugins in C++. The obvious
> solution was to translate a mud-lang to C++, then to a shared library, then
> load it in dynamically... at this point I ran amok with language issues and
> never really recovered.

  Personally I think that having a scripting language which is converted
to native C or C++ for heavily used code is the way to go; I actually 
implemented something like this on a very experimental basis.
 
> James
> 
> -- 
> MUD-Dev: Advancing an unrealised future.

-- 
	Jo

Harmony - the project to create an LGPL Qt clone
http://harmony.ruhr.de