Re: Docker images for 8.6 now default to ZTS
| From: | Rob Landers | Date: | Sat, 21 Nov 2026 14:14:20 +0000 |
| Subject: | Re: Docker images for 8.6 now default to ZTS | ||
| References: | 1 2 3 4 5 6 7 8 | Groups: | php.internals |
| Request: | Send a blank email to internals+get-133171@lists.php.net to get a copy of this message | ||
On Sat, Nov 21, 2026, at 14:34, Matthew Sewell wrote:
>
>
> On Sat, Nov 21, 2026 at 12:28 PM Rob Landers <rob@getswytch.com> wrote:
>> __
>>
>>
>> On Sat, Nov 21, 2026, at 12:26, Matthew Sewell wrote:
>>>
>>>
>>> On Sat, Nov 21, 2026 at 11:10 AM Rob Landers <rob@getswytch.com> wrote:
>>>> __
>>>>
>>>>
>>>> On Sat, Nov 21, 2026, at 11:46, Matthew Sewell wrote:
>>>>>
>>>>>
>>>>> On Sat, Nov 21, 2026 at 5:27 AM Marc Henderkes <m@pyc.ac> wrote:
>>>>>> __
>>>>>> On 21/11/2026 06:28, Benjamin Außenhofer wrote:
>>>>>>>
>>>>>>> Am 20.11.2026, 12:23:15 schrieb Matteo Beccati <php@beccati.com>:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> I've been mentioned in:
>>>>>>>>
>>>>>>>> J÷7bÆã�
>>>>>>>> AsiÊ\https://github.com/docker-library/php/issues/1719
>>>>>>>>
>>>>>>>> and I became aware that all PHP docker images now are ZTS builds.
>>>>>>>>
>>>>>>>> It was suggested to do so and apparently the Docker team took it as
>>>>>>>> an
>>>>>>>> official request. Was it?
>>>>>>>>
>>>>>>>> If not, I'm not sure it was appropriate to experiment like
>>>>>>>> this on such
>>>>>>>> a large user base, before publishing an RFC, discussing it and
>>>>>>>> voting.
>>>>>>>>
>>>>>>>> For reference:
>>>>>>>> -
>>>>>>>> https://github.com/docker-library/php/issues/1673
>>>>>>>> -
>>>>>>>> https://github.com/docker-library/php/issues/1689
>>>>>>>>
>>>>>>>>
>>>>>>>> Cheers
>>>>>>>> --
>>>>>>>> Matteo Beccati
>>>>>>> Hi Matteo,
>>>>>>>
>>>>>>> First, the Docker images are not an PHP project, so I don’t think an
>>>>>>> RFC is in the scope to change this.
>>>>>>> The „official“ part of these images is referencing that they are
>>>>>>> official docker images, not official PHP images.
>>>>>> Correct, aside from PHP not being in a position to "officially
>>>>>> request" changes to Dockers official images, I think
>>>>>> https://github.com/docker-library/php/issues/1673 should make it
>>>>>> fairly clear that the maintainers didn't understand my question as "official".
>>>>>>
>>>>>
>>>>> Hi,
>>>>>
>>>>> This caught us by surprise too. Having read through the issues I don't
>>>>> think there was any problem with #1673 but it clearly was interpreted as an official request from
>>>>> the PHP project.
>>>>>
>>>>> From the Docker maintainers, clarifying why they did it on
>>>>> https://github.com/docker-library/php/issues/1689:
>>>>>
>>>>>> PHP upstream recommended that we switch the images to use ZTS by default
>>>>>> (#1673 <https://github.com/docker-library/php/issues/1673>)
>>>>>> for 8.6 and above.
>>>>>
>>>>> The other Docker maintainer is also clear (my bolding and my added italics):
>>>>>
>>>>>> On the change itself, *I was under the impression that this change was at
>>>>>> least somewhat supported by the PHP maintainers* (and there's at least one (*that was Tim*) in
>>>>>> this very thread confirming that, at least in a personal capacity, which it's also my
>>>>>> understanding is often the best we'll get here) -- if that's inaccurate and there are PHP
>>>>>> core upstream maintainers who disagree with this change, I would definitely consider a revert with
>>>>>> haste, but I haven't seen evidence that they disagree, just that they haven't necessarily
>>>>>> explicitly agreed (which is something you're going to be hard-pressed to get, for reasons
>>>>>> described upthread).
>>>>> So it doesn't matter how it was intended, or whether (as both Benjamin and
>>>>> Gina have said) that they're official Docker images, rather than official PHP images, the
>>>>> change was made because the Docker maintainers considered the PHP project to have asked for it. I
>>>>> don't see how you can expect the Docker maintainers to understand the PHP process, so if they
>>>>> think it's an official request then that's how they'll treat it.
>>>>>
>>>>> I appreciate that a lot of the rest of the discussion on #1719 is about the
>>>>> technical merits of changing but I think this is a process question rather than a technical one.
>>>>>
>>>>> I'd consider myself a pretty well informed user, I follow the internals
>>>>> discussions, read the RFCs and check the code changes before releases. Although I use them, I
>>>>> don't follow Docker issues in the same way (maybe a flaw in our process, but from some
>>>>> discussions, would be surprised if many did). It's also mentioned in #1719 by one of the Docker
>>>>> maintainers:
>>>>>
>>>>>> We would *love* to proactively notify users of changes like this in a
>>>>>> broader way, but we don't have a good mechanism for that other than following along here in
>>>>>> GitHub (and that's been true for ~12 years, and wouldn't catch upstream changes either, so
>>>>>> it's really more of an open industry question than one specific to us).
>>>>>
>>>>> Both Daniel and Matteo have commented, either here, or on #1719, and I think
>>>>> both were also caught by surprise (and am not suggesting there is any reason why they shouldn't
>>>>> be).
>>>>>
>>>>> My concern is this is happening in the wrong order. We use Docker images in
>>>>> commercial systems. As well as the technical choices there are a number of other processes,
>>>>> integrations and documenting that we go through, and using official Docker images is always an
>>>>> easier sell, especially with security teams, than others. Whether that is right or wrong from a
>>>>> technical perspective is often irrelevant from a business perspective.
>>>>>
>>>>> We don't need ZTS at this moment for our most critical applications. If
>>>>> ZTS becomes the PHP default (via RFC) and it is clear from discussions here that is the case then I
>>>>> will use my testing and operational budget to ensure that our application works on it with no impact
>>>>> (rightly or wrongly, the clock issue would have caught us). Until that point, or a significant
>>>>> upgrade is driven from a business need then I won't.
>>>>>
>>>>> This would have (I can see tianon is working on a solution under #1722, so
>>>>> thank you), essentially then shut off official Docker images before a decision is made in the normal
>>>>> manner based on a request that has been interpreted as officially from PHP.
>>>>>
>>>>> I also think this is really shooting PHP in the foot. At the moment
>>>>> there's so much good work going on around changing security settings, and improving defaults so
>>>>> that the majority of users get a better experience.
>>>>>
>>>>> Those other discussions are happening transparently on this channel, and people
>>>>> have a single place they can reasonably follow and expect to know what's going on (even if they
>>>>> rarely comment). Even this point (from #1719):
>>>>>
>>>>>> I think that's the biggest issue, but at least for php-src
>>>>>> there's no forward compatibility promise between minor versions. E.g. 8.6 can (and will) have
>>>>>> other breaking changes than the ZTS default.
>>>>>
>>>>> There isn't, but those changes are all listed and therefore plannable, by
>>>>> following along here.
>>>>>
>>>>> Finally, I don't think this should ever be a goal driven by a change to a
>>>>> default in an official image:
>>>>>
>>>>>> That's the largest reason for why it was even suggested: see what
>>>>>> breaks while there's still a way to fix it.
>>>>>
>>>>> There's the goal that's been mentioned (and was only highlighted
>>>>> yesterday around the home page) of trying to get more corporate technical leaders to consider PHP.
>>>>> If I see that, and believe that is the case, then there is no way I'm considering this for my
>>>>> business critical system.
>>>>>
>>>>> I would ask that if someone did request/know that such a change were to happen
>>>>> again in a key other area that it was at least mentioned on this list. That would allow anyone
>>>>> interested to go away and find more detail, and hopefully stimulate conversation, even if it
>>>>> didn't need an RFC.
>>>>>
>>>>> Thanks very much for your consideration (and everyone's efforts to improve
>>>>> this and make it all work).
>>>>>
>>>>> Matt
>>>>>>
>>>>
>>>> It seems like you should send this email to the Docker image maintainers? As
>>>> pointed out, there would have been compatibility issues anyway for extensions. Just like any other
>>>> annual release.
>>>>
>>>> At the end of the day, the Image Maintainers made a decision and seem to stand by
>>>> it per their recent comments on the issue. As mentioned up thread, this list does not govern those
>>>> images or their notifications to users.
>>>>
>>>> — Rob
>>>
>>> Hi Rob,
>>>
>>> I don't agree, and that was the whole point of this. I'm also not referring
>>> to compatibility issues for extensions, which is why I said this was mainly a process email, not a
>>> technical one.
>>>
>>> The image maintainers made the decision because they considered it an official request
>>> from the PHP project. They've said so several times across those issues.
>>>
>>> This thread is surely the place to cover/clarify a technical change made because of
>>> something perceived to be an official request from the PHP project. If not, where is?
>>>
>>> Thanks,
>>>
>>> Matt
>>
>> Ultimately, the image maintainers are responsible for their own decisions.
>>
>> — Rob
>
> Thanks very much for your opinion Rob, I do find it odd though.
>
> In a hypothetical situation where someone had hijacked a core group members account and
> requested a change on behalf of PHP, and they made a set of decisions based on that, are you saying
> that you would say it shouldn't be corrected as it was the responsibility of the maintainers?
>
> To be clear, I'm not suggesting that's what happened here, but the outcome is
> similar.
>
> Best wishes,
>
> Matt
That sounds like a governance issue with their repository and procedure.
— Rob