The framework shootout
As you might know, I'm back from the 2009 Zend Conference. I have already blogged about the conference, but I intentionally forgot to talk about the framework shootout. That's because I felt the need to think about the feedback I had during and after the session, and I wanted to mature my reflection with the Symfony core team before going further into the discussion.
Instead of a classic closing keynote, the Zend Conference ended with a framework shootout. Several frameworks were represented on stage, mostly by their lead developer, and the audience were given the opportunity to ask any question they wanted. One of them was about PHP 5.3 and how each framework will make the transition to this latest and greatest PHP version.
Will Symfony 2.0 be compatible with PHP 5.2?
Both the upcoming Zend Framework and CakePHP 2.0 versions will rely on PHP 5.3. And for Symfony, I said it will still be compatible with PHP 5.2. From my point of view at the time, it would be a mistake to upgrade major frameworks to PHP 5.3 now for one main reason: major frameworks are used by many big companies, and upgrading to the latest version of a software fast in such companies is not always feasible. Sensio works for a lot of them, and we know what we are talking about. And no, some of our customers cannot "simply" compile their own version, because they rely on standard Linux distributions and their associated professional support; or because they have hundreds of machines and don't want to install a specific version of PHP just for some projects. If you attended the keynote, don't feel angry yet, keep reading.
Apparently, many people were quite disappointed with this decision, so I started to think about this matter again. I made the decision to still support PHP 5.2 for Symfony 2.0 almost a year and a half ago. At that time, it was obviously out of question to only support PHP 5.3 and I thought Symfony 2.0 was right around the corner. Time flies and Symfony 2.0 is far from being ready.
So, given that PHP 5.3 is now stable and released, and given that Symfony 2.0 is definitely not right around the corner (probably for late 2010), does it still make sense to continue supporting PHP 5.2? Or do we need to reconsider my decision?
As hinted in this post title, the Symfony core team has changed its mind, and Symfony 2.0 will leverage PHP 5.3 and drop PHP 5.2 compatibility. I hope some of you feel better now, but I can already hear some others starting to grumble. Bear with us, as this decision was not made overnight. Keep reading this post to understand the reasoning behind this decision.
Symfony 2.0 on PHP 5.3: A better language
From a technical standpoint, using PHP 5.3 for Symfony 2.0 is great news. First, because the language has evolved and sports a lot of new exciting features like namespaces, closures, anonymous functions, late static binding, SPL enhancements, better Windows support, and much more. PHP 5.3 is also the fastest PHP release ever, which is a great benefit for complex beasts like frameworks.
Symfony 2.0 on PHP 5.3: Better tools
PHP 5.3 is also about a new ecosystem. The tools are getting better. The old PEAR installer (the one used to manage symfony plugins) is being phased out and replaced with Pyrus, a modern and robust software installer. Symfony 2.0 will definitely be able to leverage this new tool and take advantage of the native phar support to provide simple and better plugin management tools.
Speaking of tools and third party libraries, let's talk about those we use natively in Symfony.
First, our beloved Doctrine ORM. As announced some time ago, Doctrine 2.0 will be only compatible with PHP 5.3. And man, Doctrine 2.0 is gorgeous. Doctrine 2.0 is one of best things that's happened to PHP in a long time.
Then, there is the Zend Framework. Symfony does not rely on the Zend Framework for its core features, but a large number of Symfony developers use some Zend Framework components in their Symfony projects (to add a search engine, connect to some well-known web services, create PDF documents on the fly, ...). Of course, using the new Zend Framework 2.0 components will require that we share the same requirements.
Using PHP 5.3 is about having a better integration with the next generation of frameworks, tools, and third-party libraries out of the box.
Symfony 2.0 on PHP 5.3: Helping the community
Last but not the least, it is also about helping the PHP community. Symfony was one of the first frameworks to join the "Go PHP5" initiative some years ago when the community tried to help people adopting PHP5 faster.
The adoption rate problem is probably what made our decision much more easier to make. As some people said on Twitter and during the keynote, if all major PHP frameworks and libraries start using PHP 5.3 for their next major version, Linux distributions, third-party tools (IDEs, ...), and companies will probably need to embrace the new version faster. And if Symfony can help in this effort, we would be very happy.
Of course, we won't change Redhat's plans overnight. They have a quite strict roadmap, and because of their long-term support, they won't upgrade to the latest PHP version anytime soon; but if some companies use one of these conservative Linux distributions, they will also be conservative with the version of the framework they use. The good news for these companies is that symfony 1.4 will be maintained for three full years (see below).
The Linux distribution problem is also mitigated by the native support of PHP 5.3 in the upcoming Zend Server. If companies are serious about PHP support, they can also switch to the Zend Server and have support for PHP 5.3 out of the box.
Symfony 2.0 on PHP 5.3: Backward compatibility
As Symfony 2.0 is about breaking backward compatibility anyway, the Symfony core team thinks that it is better to break backward compatibility once and for all and avoid yet another major compatibility break in coming years in order to finally support PHP 5.3.
Switching to PHP 5.3 means dropping PHP 5.2 support by embracing PHP 5.3 new
features like namespaces. It means that all Symfony libraries (from the MVC
framework to the Symfony components) will require some drastic changes like
renaming every class (renaming sfRequest to symfony\core\request for
instance; say good bye to the good old sf prefix!). For components we have
already published, like the dependency injection container, we will soon
create a 1.0 branch and release a 1.0 version. Then, the trunk will start a
new life and be migrated to PHP 5.3.
But we definitely don't want to loose people who won't be able to upgrade to PHP 5.3. And Symfony is in a very good position to cope with this drastic change, thanks to its Long Term Support (LTS) releases. Symfony 1.4, to be released by the end of this year, is our upcoming LTS release. As any Symfony LTS release, it will be maintained for three years by the Symfony core team. In concrete terms, developers using this version will have support until early 2013. That's plenty of time for projects which cannot afford migrating to PHP 5.3 yet. And keep in mind that the maintenance covers forward compatibility with newer PHP versions. Symfony 1.3/1.4 already works on PHP 5.3.0, and will work on upcoming 5.3.X minor releases.
Conclusion
To sum up, we have decided to take the best of both worlds:
symfony 1.4 (with support until early 2013) will be the best version for existing projects and conservative companies.
Symfony 2.0 (probably to be released late 2010) will be the best version for new projects and companies willing to either install PHP 5.3, use the Zend Server, or install the "right" Linux distribution.
I hope you will all understand and approve this move. The Symfony core team is really excited about the opportunities it gives us, and we think the Symfony 2.0 release will be a blast. Wish us luck!
Thanks to Kris Wallsmith, Fabian Lange, Jonathan Wage, Dustin Whittle, and Stefan Koopmanschap, members of the Symfony core team, who helped me to thoroughly mature my decision and for reviewing this blog post.
I agree with that decision. Hopefully, over a year from now, people would have upgraded to 5.3 anyway.
That's a good news!!! Thanks Fabien, for this effort.
I think this is a good move. The best way to iron 5.3.x into maturity is to have people use it, the sooner the better.
The decision for ZF was a hard one as well. One point that you bring up here was also considered by us carefully: how often do we want to break BC with new major versions? Our consideration was that if we waited until 3.0, we had another 2-3 years beyond a 2.0 release before we could do so. This is a disservice to those developing with the framework, and also to the greater PHP community (which often needs the backing of large projects in order to push new versions into mainstream usage).
I'm happy to hear your points, and certainly feel you're doing the right thing -- you'll get flack with either decision you make, but your points are quite solid.
That's a really great news Fabien. Symfony has to stay innovative and you choose the good option.
Regarding to symfony 1.4 LTS for existing projects, I couldn't be so enthusiastic considering the quantity of 1.0 stuff not supported on 1.4...
I also agree with that, but I'd just like to say that for me PHP 5.3 is not really a stable release and I would wait readily (and my system manager will agree) for version 5.3.1 before using it with symfony
I completely agree with this. Great move!
Fabien, I think this is the right decision. I agree with your points on the upgrade path for some large and small companies on 5.3 but like it was said in the shootout I agree that the PHP community and Open Source projects are drivers in making these upgrades happen. If no one adopted 5.3 for another 2-3 yrs we would have the same issue as we do now.
great! the php community has to move on, namespaces are very important for organizing large software systems .. i wish php had them since its first release. Beside this closures and anonymous functions can make us more productive und hopefully symfony and zend framework will allow us to do so.
This is great news. Looking forward to the upcoming releases.
I'm agree too!
You're welcome.
I just hope the switch to namespaces won't mean even more verbosity of the code.
The goal should be to keep the code simple and clean, but as far as I've seen them in action, scripts became verbose in a bad way. I just hope that this will happen only to the core symfony files, and not to the ones the developers write. That will be a huge #fail (ops, that's not twitter :) )
Great news !
I think its time to tar and feather Redhat for their idiotic policies. Note I am fine with the core of RHEL sticking to one version. They do however maintain more uptodate binaries as part of their AppStack product. However it bundles JBoss and is therefore way way way too expensive. I think its time we really make our voices heard that we want them to sell an unbundled product that just provides a scripting language platform.
That being said, looking at their previous release history (http://en.wikipedia.org/wiki/Red_Hat_Enterprise_Linux#Version_history) they are do for an immediate update, but I have heard nothing in that direction.
The talk you see around community is turning into action. This is one more sign that the PHP community is maturing by leaps and bounds.
Where is the roadmap for this LTS idea in Symfony? We knew 1.1 wasn't being supported longterm but there wasn't much said about 1.2 or 1.4 support. Is 1.4 backwards compatible to 1.2? Obviously not to 1.0 without ugliness.
Thanks for that update and the great news! I will switch our (OpenSUSE) servers to 5.3 anytime soon and I think that 1 more year to go is far beyond what anyone should need to get the latest PHP version running.
@tayhimself: http://www.symfony-project.org/blog/2009/08/06/about-symfony-1-3-and-symfony-1-4
i appreciate this! so symfony will still be the first choice in future!
@tayhimself (and others) We are cutting "a lot" to make supporting "a clean" version for 3 years. If you consider something listed there in error please discuss on the developer mailing list.
If you really need a deprecated feature, but nobody else thinks it should be supported, you are free to take the code and just use it for your apps (as a plugin?)
I'm really happy with this!!!
Though PHP 5.3 is already out, I can't use it so much as I like to, because all the projects I manage and maintain used Symfony 1.2 or a minor version, then, I' m still bound to PHP 5.2.
With this news now I'm sure I'll use Symfony 2.0 with PHP 5.3 and finally have the power of namespaces and my beloved closures :D, in my everyday work!!!
@tayhimself: I have just added information about the symfony 1.4 version on the installation page
wouldnt the namespace for request be \symfony\core\Request ?
I think this is a good call. One way apple stays so far a head is that they are able to move to the next version. They killed the floppy drive. You look at MS they can't ever let support for anything die. However in reality support for older stuff with MS products gets worse and worse.
The PHP community could learn a lot from the Ruby's community agility in embracing new versions. It makes me sick that some people still do their main dev in PHP4!!!
If you let people say I am not ready to upgrade then people will never be ready. If you tell them to be ready by a date then you are more likely to get people to follow you. If they don't want to move forward they problem aren't that helpful in the first place.
It's a very good decision as far as I'm concerned. One thing is the speed of modernisation, and the other one is compatibility with other frameworks. Reusing Zend libraries in symfony is really a GREAT feature. It would be a big pity to lose it...
It's very good decision. I was little nervous reading this but support for 1.4 is long enough to finish existing project and start thinking about upgrading them in the future. Some projects needs year or two to finish because of many reasons and need stability. I'm waiting for some missing sections in Symfony Components and of course Symfony 2. Thanks for all work.
After reading your blog post about zend conference , I was really wondering why you have not posted about your experience in framework shootout. How was the audience ? Hope you impressed the audience with symfony. Anyway good decision too.
Yeah, I second the decision to move symfony 2.0 to php 5.3. What could be better than that. We want a framework that is always looking forward (as how you guys have showed in symfony). Good job ;)
Everything will happen again with php6, perhaps we should think about it too. Zend server adoption is a good thing, does it include suhosin ?
Great call. I had my concerns about the short-term roadmap until the recent decision to extend support for 1.2 another three months. But the long-term roadmap is excellent.
Exceedingly wise.
symfony1 was recommended to be used on a dedicated server not a shared one. Will that be true for Symfony 2 as well?