]RFC] Do we want/allow "undefined symbol" error or not?

Hi all, As you might already notice my RFC for session save handler module, There is a issue how to deal with "undefined symbol" error for rather usual configuration. It seems I need to establish agreement issue by issue. Here is first one. Do we want/allow "undefined symbol" error for usual configration? I hope everyone agree, it's not acceptable. Technically, it is possible not to have "undefined symbol" errors for sub modules. (Don't you think stupid, if Windows app dies at start up and spits strange error message with rather usual usage. Don't you think it is a design flaw? ;) I'll post other issues later, see if we can agree on how to handle modules in a module. I'll take no reply as implicit "'undefined symbol' error should not be allowed for usual configuration" :) Other issues coming are: 1) header file location 2) dealing with globals and function reference 3) module initilization order control 4) location of sub module 5) other, may be. PS: I've said yes to locate save handlers to PECL at first, so I'm not against to place sub modules to other locations, if all issue is cleared. -- Yasuo Ohgaki