Re: [VOTE] 64 bit platform improvements for string length and integer

On 1/29/14, 8:16 AM, Anatol Belski wrote: > Rasmus, > > On Wed, January 29, 2014 16:37, Rasmus Lerdorf wrote: >> On 1/29/14, 7:00 AM, Anatol Belski wrote: >> >>> please look here >>> >>> >>> https://github.com/weltling/mongo-php-driver/compare/mongodb:master...m >>> aster >>> >>> two hours of work and it's about 70% of the porting, compiling with 5.5 >>> and str_size_and_int64. This is the hard caliber of course. >> >> And that is only 70%? >> >> >> This is a massive effort that will have to be done on each of the 1000+ >> extensions out there. It is going to take years before the extensions catch >> up and therefore will slow PHP 5.6 adoption to a crawl. This is not >> something that belongs in a point release. >> >> -Rasmus >> > 70% done in two hours on such a monster? The last say 5-10% might take > some longer. Say for that caliber - 3 workdays (port, fix beyond, test) to > be done is realistic. The start into porting Zend, main and other very > core was very long and painful, but doing the same with extensions now is > an exponential speedup. > > Where comes the number 1000+ from? Last year what I could see on PECL - > about 300 extensions, about 40% really active. Clear, a number is also not > in PECL, like GPL'd and so, but 1000? Massive effort is rather a big > exception. And it's not like one would keep off and amuse himself while > the exts are dying. > > Reading such stats > http://w3techs.com/technologies/details/pl-php/5/all I > wonder what is with the 5.5 adoption? If that data don't lie, despite the > amount of work done on 5.5, including PECL, it's not that wide spread as > one would wish. At least, it's being adapted slowly. Do we expect 5.6 > adoption to be any faster? > > What I'm conducting to is that - the porting concerns are very > overestimated, the adoption of 5.6 might go much slowlier as one could > wish. And the last which should be the first - the RFC solves platform > issues which retard the PHP improvement as the whole. Yahoo alone has 500+ custom extensions. Add the number of in-house extensions across every other company that has written their own plus all the ones on Github. It is easily over 1000. And it is a couple of hours work for you, the person who knows absolutely the most about this stuff. What takes you 3 hours is going to take someone else 3 weeks. They need to identify the fact that they need to make large changes. Then they have to learn what needs to be changed and find time to actually do the work and in the case of Yahoo they then have to multiply that work by 500+ as these things are written by different people in different parts of the company. And then when we release 6.0 we are likely going to ask them to do it all again. If Yahoo were to ask me if they should ever upgrade to 5.6, I would advise them to wait for 6.0 and just do this work once. And yes, I know not everyone is Yahoo. But the same applies just at a smaller scale to most sites even if they haven't written their own extensions. They are going to be blocked on some 3rd-party extension that is dragging its feet getting updated and might have decided to just skip 5.6 and wait for 6.0 before doing the work. Stuff like this tends to have a cascading effect that is really hard to predict. We have seen it in the past for trivial tiny internal API changes. -Rasmus

Attachment: [application/pgp-signature] OpenPGP digital signature signature.asc