[glob2-devel] Glob2 salvage proposal

[Stephane Magnenat]

Hi everyone,

As you all know, and despite the hard work of contributors (thanks Andrew and 
others), glob2 hasn't been progressing very fast since the original 
developers do not have much time to work on.

So here is a proposal to try to get out of this situation. This is just an 
idea I would like your comments on, so feel free to give your opinion...

As I see the situation, getting new developers into glob2 is difficult because 
the core engine is quite obscure and not really documented, the network is 
very complex and the support library (libgag) is non-standard. Furthermore, 
the synchronization requirement makes everything very sensitive. This also 
put a strong pressure on players who have to use exactly the same version.

While I can't miraculously change the situation for the core engine (but 
Andrew is putting a lot of work there), here is a proposal for the rest:

I propose to split glob2 in a server and a thin-client. The server would do 
the actual game logic (core engine) and the client will do the rendering. The 
client will connect to the server through TCP, and download game media 
(graphics, music) dynamically. Only the server will do computations. This 
should solve both the synchronization problem and the gradients limitations 
for slow computers. The client will be generic (i.e. not limited to glob2): 
you can see the client as a sort of webbrowser but with server-side layouting 
and optimized for games (binary protocol, animations support, ...). I propose 
to base the whole thing on Qt4, because it is very well done, documented, and 
provides lots of features (nice themable gui, network, reflexivity, 
serialization, several graphical backends, ...). And of course it is GPL.

I've some idea on the network protocol, but before going into technical 
details, I would like your opinion on the overall direction.

Thanks, have a nice day,

Steph