Re: Strict typing (was: Typehints)

On 11 August 2010 15:13, Zeev Suraski <zeev@zend.com> wrote: > At 15:14 11/08/2010, Richard Quadling wrote: >> >> On 11 August 2010 12:10, Zeev Suraski <zeev@zend.com> wrote: >> > We need to remove strict typing from trunk before we release anything >> > 'official' from php.net >> >> I thought "trunk" is, to some degree, the "work in progress" / >> "developers only", YMMV branch. Pretty much anything/everything in >> there is subject to change. No money back guarantees. Etc. > > Supposedly we switched to this strategy although I'm not sure why, nor I > recall any discussion about it - although I may have missed it.  We never > ever treated HEAD this way in the CVS days.  So sure, now it's called > 'trunk', but why we should deviate from our decision making processes (as > lax as they may be) because we changed version control systems is beyond me. > >> For an official release, even as a "Here is what we are working on. It >> might not be perfect, but we like it" release, a separate branch would >> be created. > > It's really not a matter of branches, trunk or HEAD.  It's a matter of what > 'php.net' puts its virtual stamp of approval on.  If 5.4 alpha 1 came out > with strict typing in it, it would send two very strong messages to the PHP > community: > > 1.  The next version of PHP is going to be named 5.4 - something that wasn't > agreed upon (although personally I don't mind that much). > 2.  "We think strict typing is a good idea, here, play with it".  Well, > turns out that the collective 'we' doesn't really think that at all.  It's > no big news either, it's been known for many months. > > That goes back to my first paragraph.  Personally, I don't like the 'shoot > first, ask questions later' approach that we supposedly switched to > recently.  To me it makes a whole lot more sense to discuss first, and only > once a decision is made - go ahead and implement it.  Whether we go formal > with RFCs or less formal on internals@ (depending on the scope) - either way > it's way better than committing first and only then discussing.  Once in > trunk we suddenly need a great reason to remove it, since trunk is now the > new 'status quo'.  Thankfully in the case of strict typing there was a > strong, clear message from the community 'don't do it', but what about > smaller features? > > 'Shoot first, ask questions later' equates 'bias for change'.  Is that > where > we want to be?  IMHO no, we should carefully consider every change we make > to the core language at this point in time. > > Maybe I'm old school, but in my opinion, trunk should only contain > agreed-upon features.  It should also always build and pass tests > successfully.  It's not the wild-west version of PHP, it's PHP's next > version, in progress.  Want to work on something experimental or > controversial?  Do that in a branch, merge it if & when it gets accepted to > the language. > > Zeev > > Thank you for that Zeev. -- Richard Quadling.