[Underware-devel] Current network packets structure

[Loic Dachary]

        Hi,

        It appeared to me today (while working with Henry, welcome
Henry ;-) that the current code wrapping network packets leads to an
extremely complex set of modifications to do something as simple as
adding a new packet type.

        Unless someone has serious objections by tomorrow morning I'll
trash the current system completly (not all the network code, just the
code used to access packets ;-). I will replace it by assuming that:

        - All packets are stored in native Python structures (dict)
        - C++ code accesses the data members of a packet using the
          Python C++ API to dict objects.
        - A minimal C++ class encapsulates the gory details and has an
          API that looks like this:

                void GetMember(string, &float)
                void GetMember(string, &int)
                ...
        
                and the same set of SetMember methods

                The constructor allocates a first class Python object and 
                INCREF it, the destructor DECREF it.

        Using Python objects instead of simple structs would be a
questionable if there was a performance issue but network code is not
really our concern in this matter. We do need, however, packets that
can be easily defined and used seamlessly in Python or C++.

        The reason for doing this now (instead of implementing
gameplay code) is that the current complexity is a significant
drawback when prototyping a new gameplay code that needs new packets
to circulate between the client and the server.

        Cheers,

-- 
Donate to FSF France online : http://rate.affero.net/fsffrance/
Loic Dachary, 12 bd Magenta, 75010 Paris. Tel: 33 1 42 45 07 97      
http://www.fsffrance.org/   http://www.dachary.org/loic/gpg.txt