Loic Dachary (2001-11-14 13:11:20 +0100) : > That plan seems safe to me. Of course one could argue that we > should use some XML bus technology since we're trying to transmit > messages to/from server and clients. Unless someone masters this > aspect we are probably going to stick to a hand made communication > and error handling between the server and the clients.
At this point, I'd like to suggest using an already existing bus technology. I am not aware of the existence of any "XML bus", but the word "bus" you mentioned whispered "CORBA" to me. Now for some anti-FUD (it might not be needed, but I've had to fight against fo much disinforation when I introduced CORBA at work that I prefer not to take any risks): I've been working with CORBA for one year and a half, and I can testify it is *not* *at* *all* heavyweight or hard to program with. It works, it's lightweight, it's interoperable, it's cross-language (and cross-architecture and cross-network etc.). I can provide help to get the thing started. This would be in the form of 20-line skeletons of servers and clients (since it's all that it takes) in both Perl and Python using ORBit, or in help on designing the interfaces (that's not much harder than an OO model). I'm told there are also bindongs for PHP, and there obviously are some for C and C++. "Sticking to hand-made communications and error handling" is a major commitment, and I would strongly advise against it. I've been involved in a few projects where different modules needed to communicate, and programming the sockets by hand, designing a protocol, and handling errors is a huge task. Roland. -- Roland Mas La menace de la baffe pèse plus lourd que la baffe elle-même. -- in Sri Raoul le petit yogi (Gaudelette)
