On Fri, Nov 12, 2004 at 03:48:18PM -0500, Michael G Schwern wrote: > On Fri, Nov 12, 2004 at 07:00:17PM -0000, Orton, Yves wrote: > > I disagree. Having modules in the core is one of the few ways to guarantee > > that the module will be uniformly available. > > Its very unlikely to get into the core. > > To summarize the (unofficial?) core policy on modules (unless something > changed while I wasn't looking), new core modules are considered if:
Looks correct to me: > 1) Its helps build or test the core. > 2) It helps installing more modules. > 3) There's no way to do it outside the core. > 4) Its needed by another core module. And this is the killer: > And then remember that anything we stick into the core is with the core. > FOREVER! Even after the module is no longer interesting nor best practice. > Not only is this dead weight on the distribution but it discourages competing, > improved implementations as the one in the core is "official". We have some regression tests that can manage to go bad and eat CPU forever. Having something to kill of rogue core tests would be useful, which this might fulfil. Except that currently it doesn't exist. So please could we just skip the whole "is it in the core or not bit" debate and get on with working code. This thread started off well, but is turning into a more heat than light thread. There will be no consensus on being in the core or not at this time. Nicholas Clark
