Re: PEAR2 standards: anything good at all?

Lukas Kahwe Smith wrote: > till wrote: > >> All in all, I think this discussion is really healthy as it drives >> things forward and makes people think. So thanks again for all the >> benchmarks and opinions here. > > Parts of it were healthy, but I think we have effectively managed to > create one of the largest threads out of something I personally find a > no brainer. I try hard to be objective here, but I find the bulk of the > commentary to be surprisingly reactionary. Its like the only thing that > sucks worse than the past is the future and today its raining too. Maybe > its time to move to Pluto. > > regards, > Lukas Ok, let's assume, we get rid of all *_once calls in all PEAR2 packages, and force a user to go with either allfiles.php (for opcode cache environment only IMO) or an implementation (user-provided or PEAR-provided) of an autoload function. Given the above, how do we implement factory() methods in PEAR2 packages, especially in the cases of fallback drivers and/or optional dependencies? A lot of factory() methods currently test existence of classes in some way or another, and return PEAR_Error instances in case of failure. In other terms, would we forbid to pass FALSE to class_exists() in such factory() methods? Obviously, if we try to load a class that doesn't exists a fatal error will occur, which may not be a desirable behavior, specially for a factory() method. Submit sample code please :-) -Philippe