|A formal release process proposal||Fabien Potencier||9/16/12 10:01 PM|
My keynote last week at Symfony Live London was about adopting a formal
release process. In fact, I've talked about adopting a shorter release
cycle for Symfony for quite some time now, and I think that this is the
right time to discuss it.
As we have all noticed, Symfony enjoys a large community of "core"
developers: a core developer being someone who contribute to Symfony on
a regular basis. The flow of pull requests has been outstanding and
steady for the past two years, and with such an activity, trying to
release often without a clear roadmap is quite difficult. Adopting a
more formal release cycle will also give more visibility to the
contributors and allow for everyone to understand when a new feature
might be available in Symfony.
So, here is my initial proposal, which is the one I've talked about
during Symfony Live and of course, it is up for discussion. I would like
to apply the new release process as soon as possible and if possible for
Symfony 2.2. And whenever we all agree on the final version of this
proposal, it will be included in the official Symfony documentation.
This release process only applies to the code hosted on the
symfony/symfony repository, but of course, I hope that third-party code
related to Symfony (like the Symfony bundles) will also adopt it (at
least, just for the timeline).
Let's list the goals for the new process:
* Shorten the release cycle;
* Keep backward compatibility as much as possible;
* Enhance the overall quality of the framework (not just the code, but
documentation, bundles, ...);
* Give more visibility to our "customers": developers using the
framework to get their job done and Open-Soure projects using/embedding
* Improve the experience of Symfony core contributors by controlling
the flow of incoming pull requests (why pull requests are not always
merged right away? when will a new feature be merged? when breaking BC
is acceptable? ...);
* Coordinate our timeline with projects that we are using (Doctrine,
Propel, Monolog, Assetic, Twig, ...) but also with projects that are
* Give time to the Symfony ecosystem to catch up with the new versions
(bundleauthors, documentation writers, translators, ...);
* Allow developers to benefit from the new features faster.
That's a lot to take care of!
So, without further ado, here is my plan.
Historically, we've been able to release a new major version every year
since 2005. Nothing was even written about that, but that's what we did.
From now on, I propose to adopt a *time-based model* for Symfony and I
think that having a new major release every six months is a good
compromise: it gives plenty of time to work on new features but it also
allows for non-ready features to be postponed to the next version
(without having to wait too much for the next cycle).
Six months should be fast enough for developers who want to work on the
latest and the greatest; but at the same time, companies might want more
time to learn and upgrade. The way to make everyone happy is to ensure
an easy upgrade path from one version to the next one. Take Twig as an
example: I've been able to release a new major version every month and a
half since 1.0; that's very fast and it has been possible because we've
kept backward compatibility between all major releases (and of course
the scope of Twig is also smaller).
Six month releases mean that two releases fit in a year and so,
everybody knows when releases will be made without having to check on
the website: for Symfony it will be at the end of May and at the end of
November of each year. That brings predictability and visibility.
The key is keeping backward compatibility. We must be much more careful
when breaking backward compatibility; and the possibility to break
backward compatibility depends on the component we are talking about.
The following components must never break backward compatibility because
they are the low-level architecture of the framework and also because so
many people rely on them:
Backward compatibility should be easy to keep for the following components:
And these components should probably become more stable soon, but that's
not that easy (yet):
Six months can be seen as a rather short period to make a new release,
especially if we look at what we did in the past. I think we can make it
work because we have now more people able to help, but also because the
six month period itself should be cut in shorter periods:
* Development: 4 months to add new features and to enhance existing ones;
* Stabilisation: 2 months to fix bugs, prepare the release, and wait
for the whole ecosystem to catch up.
During the development phase, we can revert any new feature if we think
that we won't be able to finish it in time or if we think that it won't
be stable enough to be included.
During the stabilisation phase, some developers might still work on new
features for the next version, but it would be better if most developers
can concentrate on finishing the current version.
By the way, when I have a look at the pull requests today, I think that
we already have enough features for Symfony 2.2.
Long Term Support release
We've not yet published our LTS release for Symfony2. As I mentioned it
in the past, the first LTS should be Symfony 2.3.
Each LTS release will be supported for a 3 year period but it will also
be supported for at least a year after the next LTS is released. So, it
means that we are going to release a new LTS version every two years.
This dual release cycle should make everyone happy. If you are a fast
mover, you want to work with the latest and the greatest, stick with the
standard support releases: you have a new version every six months, and
you have two months to upgrade to the next one. If you are a big
company, and you want more stability, stick with the long term support
releases: you get a new version every two years and you have a year to
To make things more concrete, here is the schedule for the next few
* Symfony 2.2 will be released at the end of February 2013;
* Symfony 2.3 (the first LTS) will be released at the end of Mai 2013
(only 3 months after 2.2 as it will be a "special" release in the sense
that we will mainly remove the 2.0 BC layer and also because I think
that May and November are the best months for releases);
* Symfony 2.4 will be released at the end of November 2013;
* Symfony 2.5 will be released at the end of Mai 2014;
So, why not releasing Symfony 2.2 earlier as we already have so many
features waiting in the pull request queue? Because of the next section:
this is our last chance to break backward compatibility.
After the release of Symfony 2.3, backward compatibility will be kept at
all cost. If it is not possible, the feature/enhancement will be
scheduled for Symfony 3.0. And the work on 3.0 will start whenever we
think that we have enough great features under our belt to make it worth it.
After Symfony 2.3, non LTS releases will be maintained for 8 months to
give people plenty of time to upgrade (keep in mind that even if no BC
breaks will have occurred, you might need to upgrade your applications
to benefit from the new features and the new best practices).
To make the new process works well (no BC and a fixed schedule), we need
to formalise the contribution process a bit more. Every new Symfony
feature or enhancement must be worked on via Git pull requests. A few
months ago, we formalised the pull request process a bit by adding a
list. But I've done a poor job in enforcing the rule. So, I'm going to
be uncompromising about it now and at the same time I'd like to
introduce even more checks in the list.
A pull request will only be merged if the following rules are met:
* The code is correct and it uses the Symfony way of doing things
(naming conventions, coding standards, ...);
* The new code is tested (or the bug to fix is covered by tests) and
all the tests pass on all supported PHP versions;
* The documentation has been updated (with a pending pull request on
* The changelog and upgrade files have been updated;
* No backward compatibility break has been introduced;
* If it is a fix, it has been applied to the oldest and still
supported Symfony version;
* For major features, a RFC has been written, discussed, and approved.
As I said at the beginning, this is a draft, and you are all welcome to
chime in and propose changes.
Sensio CEO - Symfony lead developer
sensiolabs.com | symfony.com | fabien.potencier.org
Tél: +33 1 40 99 80 80
|Re: [symfony-devs] A formal release process proposal||Javier Eguiluz||9/16/12 11:42 PM|
Fabien, thank you for sharing the proposal of the new realease process and for making it debatable.
In my opinion, the proposed changes are great and they will improve the quality of Symfony and its ecosystem.
The only drawback I see is the new set of pull request rules. I think that sticking to those rules would be awesome ... but a bit unrealistic. Documenting every change (even if you don't know if it's going to be accepted), updating changelog and UPGRADE, adding tests for all supported PHP versions, etc. for every single code change is so cumbersome that most people don't do it even for their own projects.
|Re: A formal release process proposal||David Rekowski||9/16/12 11:06 PM|
Sounds like a healthy release plan.
Two questions, though:
1) for clarification: It looks like the versioning scheme matches the
"Semantic Versioning" model (http://semver.org). Can you explicitly
confirm this? If not, how do you intend to differ?
2) Will there be a roadmap outlining the planned features for releases
conveniently, without having to sort through the issue tracker and
On Sep 17, 7:01 am, Fabien Potencier <fabien.potenc...@symfony-
|Re: [symfony-devs] A formal release process proposal||Fabien Potencier||9/16/12 11:54 PM|
On 9/17/12 8:41 AM, Javier Eguiluz wrote:Oh, I forgot to mention this aspect of things. Of course, we will not
ask all pull requests to be merge-ready from the get go. But whenever a
PR will be merged (the feature is accepted and the way it is coded is
good), then, we must have everything I've described before it can be
actually merged and only when it makes sense. When fixing a typo for
instance, the PR itself is enough of course.
> Javier Eguiluz
> www.symfony.es <http://www.symfony.es>
> <http://symfony.com/doc/current/contributing/code/patches.html#make-a-pull-request)/check>> sensiolabs.com <http://sensiolabs.com> | symfony.com
> <http://symfony.com> | fabien.potencier.org
> Tél: +33 1 40 99 80 80 <tel:%2B33%201%2040%2099%2080%2080>
>> it to security at symfony-project.com <http://symfony-project.com>
> To unsubscribe from this group, send email to> symfony-devs+unsubscribe@__googlegroups.com
> For more options, visit this group at> http://groups.google.com/__group/symfony-devs?hl=en
|Re: A formal release process proposal||Florin Patan||9/16/12 11:58 PM|
Like Javier said, one of the problems that I see would be people needing to make the changes for the documentation/upgrade files even if their PR might not be accepted in the end.
Maybe having some sort of two steps process where first a PR is accepted for merge then the rest of the things are done to statisfy the checklist are done would help people in this process. This way even if the author of the PR becomes unavailable someone else could do the rest of the things if the PR is really good and should be merged asap :)
Also a roadmap of features would be very helpful. Think for example the caching component/bundle which is now missing from the framework entierly, or support for UTF-8 routes and so on. Instead of having all people asking for all sorts of features, having a central place where they could check the future featutes/changes and have a way to either submit ideas or comments for a certain feature would definately go a long way.
|Re: [symfony-devs] Re: A formal release process proposal||Fabien Potencier||9/16/12 11:59 PM|
On 9/17/12 8:06 AM, David R wrote:Yes, I confirm. And by the way, we have been using this versioning
scheme since 2005. Nothing new here.
I'm not sure about this one. Symfony grows organically, because people
want to solve problems. So, all the new features come from developers.
We do not plan in advance, except for postponed features, or features
asked in the tracker without an implementation, or big features that
take more than one iteration to finish... which is probably enough.
|Re: [symfony-devs] A formal release process proposal||Michał Piotrowski||9/17/12 12:04 AM|
I think that what you suggest is a very good plan. If no one comes up
with a better solutions it should be put into practice.
Do you have any thoughts about new major PHP releases? Current minimum
PHP version is 5.3.3. PHP 5.4 is not adapted by enterprise systems yet
- but it is only a matter of time.
|Re: [symfony-devs] A formal release process proposal||Lukas Kahwe Smith||9/17/12 12:32 AM|
As you point out, this is infact more or less what we have agreed upon during meetings before we went to 2.0. We just failed to execute on that, probably also because we never formally wrote it down.
I think this is a very important point.
This way we do not need to do complex coordinations with releases, because the general timeframe (maybe +- 3 weeks) is very clear. However we will also need to setup some coordination for point releases (especially security stuff), but also general bug fixing. However that is beyond of the scope of this proposal.
Just to expand on why this makes sense to me also: If you want to get a feature into Symfony2, you can either get it into the next release which should be less than 6 months away .. or if its too late in the release cycle (which means its likely less than 3 months to the next release and its a big complex feature), then you just have to wait less than 9 months to get it into the release after that. So as a result nothing has really changed in how fast you can get a feature in. The difference is that you can very quickly know the point in time when the feature will make it in.
I would actually prefer April, October. May seems too close to the summer and November too close to the "lets get some last features done on this years budget". Then again one could say such a feature is the upgrade to the latest version.
We should survey our user base of large projects if 3 years is sufficiently long for them. Which brings me back to the point of needing some way to coordinate stuff with them.
sounds good to me .. though like I said I would prefer us to start in April.
Yes, we should tag such things clearly in the ticket system.
Hmm I wonder if we should make it possible for people to skip one non LTS release.
this sounds like a good idea, but for that we need to work on a better work process. Right now all the work for most PRs sit with one person. If we expect that one person to do all of the above, I think we will end up in a situation where people wait for others to do the PR even if the idea has been laid out in a ticket or the mailinglist, or PRs sit open for a long time.
I have said this before on the list: Its possible to send PRs to the repo on which a PR is based. Also its possible to create a new PR based on the repo from which the original PR is based. So please use this to assist people to finish PRs. In the same way we then also need people to help on the related doc PR.
|Re: [symfony-devs] A formal release process proposal||Lukas Kahwe Smith||9/17/12 12:35 AM|
I think we should have a few high profile features picked out. I also think we should have someone assigned to help coordinate things. Their job is to also keep a close eye on features that were merged but end up requiring further PRs to stabilize to ensure that if we do need to revert something, we do it in time etc. Their job to some extend would also be to ensure that if a PR that provides a feature we wanted in gets stuck to try and see if he can get some resources in the community. This is kind of the job of a release manager inside PHP (though RMs in PHP usually also have more technical tasks, which I dont think we need for Symfony).
|Re: A formal release process proposal||Sven Paulus||9/17/12 3:39 AM|
thank you for the symfony release roadmap!
I have one little question concerning non LTS releases:
Is this 8 months after the initial release of a non LTS release or 8 months after the subsequent release?
In case of the first alternative that would mean that we'd have to upgrade to the next release within two months, which is quite a short time regarding application release cycles in some environments (and a general hesitation to jump on x.y.0 releases ... ).
|Re: [symfony-devs] A formal release process proposal||dbu||9/17/12 7:17 AM|
> Do you have any thoughts about new major PHP releases? Current minimumi guess 5.3 will stay around for quite a while. not all hosters are
quick to update.
making real use of 5.4 means breaking backwards compatibility in a very
i think php > 5.3 is a discussion for symfony 3, not for any 2.x version.
ps: +1 for fabiens proposal from me.
Liip AG // Agile Web Development // T +41 26 422 25 11
CH-1700 Fribourg // PGP 0xA581808B // www.liip.ch
|Re: [symfony-devs] Re: A formal release process proposal||Fabien Potencier||9/17/12 7:50 AM|
Yes, you have 2 months to upgrade. But this is really your choice: if
you cannot be fast, just use the LTS releases.
This is already quite an improvement over the current situation, where
you had no overlap between releases (talking about symfony1 here).
|RE: [symfony-devs] A formal release process proposal||Michael C (UKB)||9/17/12 8:53 AM|
I personally (from experience) have found that features blocking a release can end in disaster and can upset the whole principal of a time-based release model as the feature can end up delaying the release; especially if it’s a feature that many want but not many want to help with.
As for > 5.3, at the moment 5.2 is the most used PHP version; followed closely by 5.3. However the amount of users/hosts using 5.4 is currently quite small and isn't worth thinking about until at least 2.7 or 3.0 (2015 or later). The fact that 5.3 was first released in 2009 and the fact that it still is not as used as 5.2 (2006) is shocking.
I think having to complete the whole checklist done for something straight away for something that might not be used might put off new contributors. As has been suggested a 2-step PR process where after it is mostly done; if it looks like it will be merged then complete the checklist (documentation etc.) rather than straight away. This also means that if the developer is asked to change anything significant this can be done before documentation etc. is written and it therefore means that does not also need updating.
+1 to the proposal generally though. It seems like a great way to go.
> -----Original Message-----
> de...@googlegroups.com] On Behalf Of Fabien Potencier
> Sent: 17 September 2012 06:01
> Subject: [symfony-devs] A formal release process proposal
> To post to this group, send email to email@example.com
|Re: [symfony-devs] A formal release process proposal||Christophe COEVOET||9/17/12 8:57 AM|
Le 17/09/2012 17:51, Michael C a écrit :
Merging the PR before the documentation PR is done is a bad idea IMO. It will lead to the same issue than currently: no doc PR done in many cases.
However, as Fabien said in a previous mail, having the full checklist is not required to open the PR. It would be required to merge it, meaning you can start writing the doc once the code review has validated the way it is done.
-- Christophe | Stof
|Re: A formal release process proposal||jB!nfo||9/17/12 9:29 AM|
+2 for the proposal;
Yaml component is also used by many third-party code, i think it will be good to place it with the low level components;
I also find the idea of Lukas Smith is good, release manager profile :)
finally, is there is a plane for translating the documentation ?
|Re: [symfony-devs] Re: A formal release process proposal||Christophe COEVOET||9/17/12 9:54 AM|
Le 17/09/2012 18:29, jB!nfo a �crit :
There is already some work on translations. And the official website
already displays the Italian translation (since several months) and the
French one (since a few weeks).
Other translations are also in progress but are not yet completed and so
not rendered on the official site (outdated or incomplete translations
are worse than no translation). You can have more informations in the
Christophe | Stof
|Re: [symfony-devs] A formal release process proposal||Florin Patan||9/17/12 10:49 AM|
@stof it's true that most of the current PR don't provide the documentation for it but on the other hand having a two-step good-to-merge process would ease up the problems with it.
First and foremost the PR should reach a state as close to final as possible in terms of features and code then it should be flagged as good for merge (or something that reflects the state better).
Then the PR should cover all the other aspects before needing to be merged like CS and documentation, or anything else that's decided to be done in this step. When the second step is done the PR should be changed to ready for merge and be merged as soon as possible.
This should be done for every type of PR, feature addition, feature change and bug fix. This way if for example one does the code for the PR but becomes unavailable then someone else could pickup the documentation/CS for it and do it so that the PR can be merged. I believe that people that would want a certain PR to be merged would then be motived to finish the reminder of the process for a PR to be accepted if the need occurs.
Also, if we'd have a roadmap with the changes planned for each release that could be interactive, vote for a certain feature to be done faster that the other and so on then people would know when to expect a certain thing to become available. For example see the way YouTrack from JetBrains allows it, see http://youtrack.jetbrains.com/issues/WI for PhpStorm/WI as an example (but not as yet another issue tracker), which I believe it's a very good way to keep track of what users want merged/implemented faster.
I know that Symfony is based on a community of contributors, it's a different model from JetBrains' one, so maybe voting for features wouldn't seem so obvious but in the long run contributors would be able to channel their efforts better imo.
Also validating the CS on the second step would allow one to focus on doing the functionality first then fix any quirks remaining in the code would ease up the review process of the PR, where possible.
And a well defined checklist for a PR to cover would also help keeping track of things fresh contributors need to handle to know that their PR can be merged or not.
|Re: [symfony-devs] A formal release process proposal||Jérôme Vieilledent||9/17/12 2:23 AM|
I strongly +1 this :-).
2012/9/17 Fabien Potencier <fabien.p...@symfony-project.com>
--To post to this group, send email to symfon...@googlegroups.com
|Re: [symfony-devs] A formal release process proposal||Daniel Kucharski||9/16/12 11:57 PM|
The release approach seems indeed good way to move forward. However if you would want to have symfony2 being used in a more enterprise enviroment I'm afraid that 3 years of LTS isn't enough. Especially because they are not always in the driving seat to keep up with releases. I know that for most web projects 3 years a long time, but it isn't in the enterprise world. Especially if already takes sometimes one year to deliver a project.
|Re: [symfony-devs] A formal release process proposal||Michał Piotrowski||9/17/12 11:38 PM|
2012/9/17 Daniel Kucharski <insp...@gmail.com>:
> Hey Fabien,It's true. Enterprise is a specific world.
Do you think that there is a demand for a distribution with longer time support?
|Re: [symfony-devs] A formal release process proposal||Michał Piotrowski||9/17/12 11:43 PM|
2012/9/18 Michał Piotrowski <mkk...@gmail.com>:
> Hi,I mean paid distribution with paid support.
|Re: [symfony-devs] A formal release process proposal||Nikita Nefedov||9/17/12 11:57 PM|
It will be really good if Symfony project will use youtrack (it's free for open-source projects)
|Re: [symfony-devs] A formal release process proposal||Larry Garfield||9/18/12 8:58 AM|
On 9/17/12 10:57 AM, Christophe COEVOET wrote:This is actually one of the reasons why Drupal's inline Docblocks are so
much more extensive than Symfony's. We rely on the docblocks for much
of our developer-facing documentation. That means there's no separate
patch/PR for documentation about a system, at least at the code level.
If the documentation isn't "ready", it's obvious from the PR itself.
Architectural-level documentation is a separate matter and that can't
really be merged into the code base sanely, but having more of the
documentation in the code itself makes it easier to enforce adding code
changes and matching documentation at the same time.
|Re: [symfony-devs] A formal release process proposal||Larry Garfield||9/18/12 8:53 AM|
On 9/17/12 2:32 AM, Lukas Kahwe Smith wrote:Related logistical note: Drupal 8 is currently looking at code freeze 1
April next year. However, I think it's a fair bet that we'll want to
ship on 2.3 LTS, not on 2.2, which would go out of support well-before
Drupal 8 does. We'll need to coordinate between Fabien (and whoever
else is relevant) on the Symfony side and Dries/catch (the Drupal 8
leads) on the Drupal side to see how we want to handle that. Since
we'll likely be chasing master until then we may be able to just make an
exception to our code freeze for Symfony 2.3.
No action needed right now; I'm just flagging it as a discussion point.
I'll try looping in the Drupalers on my side. I cannot speak for the
release schedule of other Symfony-using projects.
|Re: [symfony-devs] A formal release process proposal||Paul Nguyễn||9/19/12 2:28 AM|
Big thanks Fabien
Sent from my iPhone
On 17-09-2012, at 12:01, Fabien Potencier
|Re: A formal release process proposal||Florin Patan||9/19/12 2:09 PM|
What about the release of separate components / bundles shipped with Symfony Standard? How would that be handled?
Currently if a component needs an upgrade, say for a security patch, then a new release needs to be done, if I'm right, which basically goes against the idea of having the framework as a group of components glued together.
Maybe a solution like patch-levels, I think composer supports a version like : x.y.z.a and if needed add the changes to composer to support this type of updating as well?
|Re: [symfony-devs] Re: A formal release process proposal||Christophe COEVOET||9/19/12 2:19 PM|
Le 19/09/2012 23:09, Florin Patan a �crit :
> What about the release of separate components / bundles shipped withcomposer supports 4 level of numbers for the version number.
But anyway, if people are using composer, update from a maintenance
release to the next one is simply a matter of running "composer update"
without changing anything else, and security patches should also be
included in a maintenance release for the current branch, not only in
the next .y release.
Christophe | Stof
|Re: [symfony-devs] Re: A formal release process proposal||Florin Patan||9/19/12 2:40 PM|
I was thinking to upgrade to a .a revision level rather that release a maintenance version sooner that planned/wanted.
Think for example if something appears in the YAML component and needs to be patched asap but at the same time you want to wait for a certain event to happen before releasing the maintenance version, like a work in patch for Process component.
Something like Ubuntu has for PHP for example. They ship a version of PHP, current is 5.3.10 but then they have the patch level so that they can still ship updates to that version without upgrading the PHP smaller increment version and thus creating confusion among users. This might not be the case for now but when going for LTS would you be doing small maintenance release as soon as needed with the whole framework or just bump the component needed?
On Thursday, September 20, 2012 12:19:41 AM UTC+3, Christophe COEVOET wrote:
Le 19/09/2012 23:09, Florin Patan a ï¿½crit :
|Re: [symfony-devs] A formal release process proposal||Larry Garfield||9/19/12 4:14 PM|
On 9/18/12 10:53 AM, Larry Garfield wrote:Confirmed that there's no issue on the Drupal end with this plan. We'll
chase master basically until 2.3 is released, then pin to that for
The caveat is that Drupal 8 is likely to remain in support for more than
3 years, so we're not sure if we should be chasing 2.x releases even if
they're non-LTS, see about a longer 2.3 lifetime, or what.
|Re: [symfony-devs] A formal release process proposal||Fabien Potencier||9/19/12 10:56 PM|
What is your support commitment with Drupal 8?
> --Larry Garfield
|Re: [symfony-devs] Re: A formal release process proposal||Fabien Potencier||9/19/12 10:58 PM|
On 9/19/12 11:40 PM, Florin Patan wrote:If the YAML or the Process component need a patch to be released ASAP,
we just release the whole Symfony package. That's what we've been doing
for quite some time now and it works well.
|Re: [symfony-devs] A formal release process proposal||Larry Garfield||9/19/12 11:09 PM|
On 09/20/2012 12:56 AM, Fabien Potencier wrote:
>>> Related logistical note: Drupal 8 is currently looking at code freeze 1Drupal supports the current and previous stable, so Drupal 8 will be
supported until Drupal 10 is released. The timeline is not fixed,
however, as Drupal operates on a "when it's ready" schedule. Until
recently the release cycle was about yearly (give or take), but Drupal 7
was a 3 year dev cycle and Drupal 8 is on track to be a 2.5 year cycle.
There are many voices calling for shorter release cycles (with good
reason), but for planning purposes I'd say we should be thinking of a 2
year cycle each for Drupal 9 and 10, so 4 years before Drupal 8 is
retired is a reasonable estimate.
(Drupal allows for overhauls and API changes in each major version,
which is why a shorter, tighter release cycle is not currently
feasible. Whether or not that will change in the future is anyone's guess.)
|Re: [symfony-devs] Re: A formal release process proposal||Thomas Lundquist||9/19/12 11:31 PM|
On Wed, Sep 19, 2012 at 11:19:37PM +0200, Christophe COEVOET wrote:Sysadmins and even some developers are a bit vary when there are a new release
of something and it have to get into production, maybe ASAP because of
a security issue. They are and should be conservative and causcious.
Having the components decoupled makes the scope for an update alot smaller
and easier to handle. And it will make it seaier to push out fixes just
for that component alot simpler.
|Re: A formal release process proposal||Daniel Kucharski||9/20/12 6:37 AM|
In my opinion a LTS lower then 3 years might be a deal for more enterprise customers which would want to use Symfony2 for a backbone/backend application (I'm not talking here about a simple marketing application). I live more in the enterprise consulting sphere (SAP business consulting) and noticing that LTS support is very important for a service delivery manager deciding to go for a given platform / tooling. Especially if the project (from blueprint to go-live) spans more then 1,5 year. A suggestion I would make is to at least offer a paid eLTS support service (extended LTS support) at some point in time. Even just having it on paper would be enough to convince managers to go for Sf2. I'm sure enterprise companies will be happy to pay for this. For instance, an eLTS of 2 - 4 years would be good.
Six month releases mean that two releases fit in a year and so,
|Re: [symfony-devs] A formal release process proposal||Lukas Kahwe Smith||9/20/12 11:15 PM|
how are we going to conclude this discussion?
Lukas Kahwe Smith
|Re: [symfony-devs] A formal release process proposal||Fabien Potencier||9/20/12 11:45 PM|
On 9/21/12 8:15 AM, Lukas Kahwe Smith wrote:As this is no small decision, I don't want to rush things out. I want to
give plenty of time for people to digest and give feedback. But as
nothing major has emerged from the discussion, I will write a pull
request for the documentation next week, wait a bit to gather more
feedback there, and put in into action just after that.
There is one thing that I have not mentioned in the discussion and I
don't remember if I mentioned it during my talk at Symfony Live London,
but we need to discuss the end of life of Symfony 2.0. If we apply what
we are discussing here, Symfony 2.0 maintenance would stop at the end of
October 2012 (2 months after the release of 2.1.0). There are two things
to keep in mind as 2.0 was a bit special:
* the form framework did change quite a lot in 2.1 and so the
transition for forms from 2.0 to 2.1 is not easy;
* merging 2.0 into 2.1 is a major PITA (just for me actually) as all
the tests were moved in 2.1, and because the form framework evolved a
lot as well, so it takes me more and more time to merge 2.0 into 2.1;
* we can decide to keep the 2 months period but still accept security
fixes for a longer period of time (6 months).
So, what do we do with Symfony 2.0?
|Re: [symfony-devs] A formal release process proposal||Thomas Lundquist||9/21/12 12:29 AM|
On Fri, Sep 21, 2012 at 08:45:19AM +0200, Fabien Potencier wrote:I do have a slight feeling quite many out there would want to keep their 2.0
running until there was an available LTS and *then* do the merging and
So my vote? 2-3 months after LTS.
You could argue that they should move alot faster but 2.1 is still
something in-between and there might be breaks. (Or did we manage to put
all BC breaks and API changes into 2.1, aka upgrading from 2.1 to LTS will be
|Re: [symfony-devs] A formal release process proposal||Lukas Kahwe Smith||9/22/12 12:20 AM|
On 21.09.2012, at 08:45, Fabien Potencier <fabien.p...@symfony-project.com> wrote:we talked about changing the merging policy in Denver. ie. we would merge bug fixes to master and then merge it down ro previous releases. i dont remember the pros and cons we listes back then. but it might become more necessary to enable delegation of responsibility.
|Re: [symfony-devs] A formal release process proposal||Christophe COEVOET||9/22/12 4:44 AM|
Le 22/09/2012 09:19, Lukas Kahwe Smith a �crit :
The big issue with this workflow is that it forces to use cherry-pick
for older branches instead of merging them together. So changes are
applied in the maintenance branch but with a different commit hash.
This is what Doctrine does and it makes it impossible to ask git which
commits have been backported (a diff will show the commits being new in
master even if they have a cherry-picked version). It requires doing the
history diff by hand each time.
Christophe | Stof
|Re: [symfony-devs] A formal release process proposal||Jérôme Vieilledent||9/22/12 4:51 AM|
That's the workflow we use for eZ Publish EE and it's perfectly fine :-)
Le 22 sept. 2012 à 13:44, Christophe COEVOET <st...@notk.org> a écrit :
|Re: [symfony-devs] A formal release process proposal||Fabien Potencier||9/22/12 6:24 AM|
I think the current worflow is the best one as we can check very easily
which patches/PRs are merged in which branches. So, I don't think that
changing this strategy would be good.
|Re: [symfony-devs] A formal release process proposal||Larry Garfield||9/22/12 8:53 PM|
Drupal does the "manual apply to each branch, newest first" approach as
well. It's not perfect but it's worked for 11+ years and counting.
> J�r�me Vieilledent
|Re: [symfony-devs] A formal release process proposal||Lukas Kahwe Smith||9/23/12 12:10 PM|
yes .. but if the current approach means everything for all branches needs to go through you, then we have a scalability issue. even if some things might be slowed down but then having the option to delegate work to multiple people, we might still have a higher "through put".
|Re: [symfony-devs] A formal release process proposal||Fabien Potencier||9/27/12 5:39 AM|
|Re: [symfony-devs] A formal release process proposal||Wouter J||10/9/12 2:03 PM|
This pull request is merged in 2 days ago and the formal release process is now official: http://symfony.com/blog/a-new-release-process-for-symfony
That concludes this discussion.
- Wouter J
Op donderdag 27 september 2012 14:39:23 UTC+2 schreef Fabien Potencier het volgende: