Re: zend module patch

Zeev Suraski wrote: > > At 23:47 10-10-01, Stig S. Bakken wrote: > > > We already broke the API numerous times in 4.0.7. The thing we didn't > > > break is the narrow binary compatibility of the module_entry structure. > > > >If this isn't a reason for 4.1 I don't know one. > > Stig, really, don't do this to me. I'm seriously considering quitting PHP > completely now :) > > I'll give you another reason which I guess you forgot - and that's the > register_globals issue. 4.0.7 and 4.1.0 are supposed to be released > simultaneously (or almost simultaneously), with the only real difference > being the register_globals default. As you can look up in the archives, I > believe it's a pretty crucial step in the deprecation of register_globals > in a user-friendly way. Okay, here's a plan (Zeev and I discussed this on IRC): 1. Change the 4.0.7 release to 4.1.0 and include the module patch. Reasoning: better to have one than two releases breaking compatibility. The intention of this patch is to end or at least reduce such compatibility problems. And an API change as big as this mandates the use of that middle digit we're all so afraid of using. :-) Sascha: apart from being an unspeakable kludge, won't having two module structs just prolong the pain of this change? Wouldn't it be sufficient to put a visible warning about it in the release notes and attach a README file on converting modules? I think people will accept that when it means an end to BC problems if more such changes occur in the future. 2. Postpone the register_globals default change to 4.2.0 3. Schedule 4.2.0 as the next release after 4.1.0 (don't even go through 4.1.1 unless absolutely necessary) 4. Either stick to the RC scheme (bug fixes only), or rename them to betas. - Stig