Disabling reflection-free Jackson serializers by default (again)?

70 views
Skip to first unread message

Guillaume Smet

unread,
Aug 17, 2026, 5:48:38 AM (3 days ago) Aug 17
to Quarkus Development mailing list
Hi,

I have been thinking about this for a while now, seeing the number of issues that get open for this feature. It comes late because I respected Mario's work too much but I think it's time for me to write this email.

We sure improved it and made it more reliable by enabling it by default but I'm really not sure the overall approach is solid enough for it to be the default.
We have seen a loooot of corner cases (which got fixed quickly but...) and we still got some reports lately. And I have no idea if we will see the end of these reports and if new Jackson versions won't come with their own challenges.

My recommendation would be to make it an opt-in feature again and go back to standard Jackson by default.
For 3.x, for sure, especially with the LTS coming, and tbh, I think I would recommend it for 4.x too.

If the approach was minimal and transparent, I would be all for it being the default, but time has shown that we have to reimplement a lot of corner cases and I'm not convinced it's the right approach for the long term.

Thoughts?

-- 
Guillaume

George Gastaldi

unread,
Aug 17, 2026, 5:57:48 AM (3 days ago) Aug 17
to Quarkus Development mailing list
+1, that makes sense

--
You received this message because you are subscribed to the Google Groups "Quarkus Development mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to quarkus-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/quarkus-dev/CALt0%2Bo-4%3D%2B24NJ6Oq%2BQHOeMQYkTLNe13O%2BBYMpppVnPovOKQAQ%40mail.gmail.com.

Mario Fusco

unread,
Aug 17, 2026, 7:06:14 AM (3 days ago) Aug 17
to quark...@googlegroups.com
Hi Guillaume,

I'm not strongly opinionated about this and I will leave this decision to you and others entitled to make it. 

I totally agree that I grossly underestimated the complexity and number of possible corner cases of this feature... and If I hadn't, I probably wouldn't have started implementing it at all. That said let me add a couple of data points:

- We had several (actually a lot of) bug reports in the first weeks after this feature has been enabled by default, but ALL those bugs have been fixed within 24 hours or less.
- It is at least 2 weeks that we don't receive a new bug report on that area, so, after a transitory period, hopefully it is much more stable now (but this could be also simply related with the fact that in the first half of August many are on vacation).

Just my 2c,
Mario

Guillaume Smet

unread,
Aug 17, 2026, 7:11:54 AM (3 days ago) Aug 17
to quark...@googlegroups.com
Yeah, for your second point, that's what prevented me from sending this email before - and I totally agree you fixed the issues at lightning speed. I thought maybe we have implemented everything now. But, even so, I think the approach is inherently brittle as there are too many corner cases.

Also what triggered me sending this email is these two that were opened recently:

I'll wait for Georgios to come back as I'm always a bit more pessimistic than he is :).

Martin Kouba

unread,
Aug 17, 2026, 7:33:49 AM (3 days ago) Aug 17
to quark...@googlegroups.com, Guillaume Smet
On 8/17/26 13:11, Guillaume Smet wrote:
> Yeah, for your second point, that's what prevented me from sending this
> email before - and I totally agree you fixed the issues at lightning
> speed. I thought maybe we have implemented everything now. But, even so,
> I think the approach is inherently brittle as there are too many corner
> cases.

There's one problem with opt-in features though - usually you don't get
enough feedback (esp. around corner cases) to stabilize it.

That said, I'm not an expert in this area, so I have no idea if it
should be enabled by default or not.

>
> Also what triggered me sending this email is these two that were opened
> recently:
> - https://github.com/quarkusio/quarkus/pull/56050 <https://github.com/
> quarkusio/quarkus/pull/56050>
> - https://github.com/quarkusio/quarkus/pull/55984 <https://github.com/
> quarkusio/quarkus/pull/55984>
>
> I'll wait for Georgios to come back as I'm always a bit more pessimistic
> than he is :).
>
> On Mon, Aug 17, 2026 at 1:06 PM 'Mario Fusco' via Quarkus Development
> mailing list <quark...@googlegroups.com <mailto:quarkus-
> d...@googlegroups.com>> wrote:
>
> Hi Guillaume,
>
> I'm not strongly opinionated about this and I will leave this
> decision to you and others entitled to make it.
>
> I totally agree that I grossly underestimated the complexity and
> number of possible corner cases of this feature... and If I hadn't,
> I probably wouldn't have started implementing it at all. That said
> let me add a couple of data points:
>
> - We had several (actually a lot of) bug reports in the first weeks
> after this feature has been enabled by default, but ALL those bugs
> have been fixed within 24 hours or less.
> - It is at least 2 weeks that we don't receive a new bug report on
> that area, so, after a transitory period, hopefully it is much more
> stable now (but this could be also simply related with the fact that
> in the first half of August many are on vacation).
>
> Just my 2c,
> Mario
>
> On Mon, Aug 17, 2026 at 11:57 AM George Gastaldi
> <gegas...@gmail.com <mailto:gegas...@gmail.com>> wrote:
>
> +1, that makes sense
>
> Em seg., 17 de ago. de 2026, 06:48, Guillaume Smet
> <guillau...@gmail.com <mailto:guillau...@gmail.com>>
> dev+uns...@googlegroups.com <mailto:quarkus-
> dev+uns...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/
> msgid/quarkus-dev/
> CALt0%2Bo-4%3D%2B24NJ6Oq%2BQHOeMQYkTLNe13O%2BBYMpppVnPovOKQAQ%40mail.gmail.com <https://groups.google.com/d/msgid/quarkus-dev/CALt0%2Bo-4%3D%2B24NJ6Oq%2BQHOeMQYkTLNe13O%2BBYMpppVnPovOKQAQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.
>
> --
> You received this message because you are subscribed to the
> Google Groups "Quarkus Development mailing list" group.
> To unsubscribe from this group and stop receiving emails from
> it, send an email to quarkus-dev...@googlegroups.com
> <mailto:quarkus-dev...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/
> quarkus-dev/CAK6y9v0ydhxOEncP6O%2B6VnLnbuzW9zSkWvdKtZhsz-
> YkPc5gBQ%40mail.gmail.com <https://groups.google.com/d/msgid/
> quarkus-dev/CAK6y9v0ydhxOEncP6O%2B6VnLnbuzW9zSkWvdKtZhsz-
> YkPc5gBQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.
>
> --
> You received this message because you are subscribed to the Google
> Groups "Quarkus Development mailing list" group.
> To unsubscribe from this group and stop receiving emails from it,
> send an email to quarkus-dev...@googlegroups.com
> <mailto:quarkus-dev...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/
> quarkus-dev/
> CALO%3D_Z72AXCmMwNrbM2U6R1TYs3z%3D7N2Sq_RQcQE_FcNZLX%2Bng%40mail.gmail.com <https://groups.google.com/d/msgid/quarkus-dev/CALO%3D_Z72AXCmMwNrbM2U6R1TYs3z%3D7N2Sq_RQcQE_FcNZLX%2Bng%40mail.gmail.com?utm_medium=email&utm_source=footer>.
>
> --
> You received this message because you are subscribed to the Google
> Groups "Quarkus Development mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to quarkus-dev...@googlegroups.com <mailto:quarkus-
> dev+uns...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/quarkus-
> dev/
> CALt0%2Bo9oqF2ymR%3DRGZmBTQxhNQXUEqJu%3DWZurtt1V2ndQjnHhA%40mail.gmail.com <https://groups.google.com/d/msgid/quarkus-dev/CALt0%2Bo9oqF2ymR%3DRGZmBTQxhNQXUEqJu%3DWZurtt1V2ndQjnHhA%40mail.gmail.com?utm_medium=email&utm_source=footer>.

Guillaume Smet

unread,
Aug 17, 2026, 7:44:09 AM (3 days ago) Aug 17
to Martin Kouba, quark...@googlegroups.com
On Mon, Aug 17, 2026 at 1:33 PM Martin Kouba <mko...@redhat.com> wrote:
There's one problem with opt-in features though - usually you don't get
enough feedback (esp. around corner cases) to stabilize it.

That I can agree with, and that's why we tried.
And Mario improved things a lot.

But the problem for me is that there were (are?) MANY corner cases - and I would have preferred an approach where we don't have that many.

-- 
Guillaume 

Martin Kouba

unread,
Aug 17, 2026, 7:57:35 AM (3 days ago) Aug 17
to Guillaume Smet, quark...@googlegroups.com
Yes, I understand. It's a pity we can't use something like gradual
rollout for extension features ;-)

>
> --
> Guillaume

Guillaume Smet

unread,
Aug 18, 2026, 3:16:05 AM (2 days ago) Aug 18
to Martin Kouba, quark...@googlegroups.com

Which could cause some data loss if you're not careful.

So yeah, I think my opinion is pretty much made, at least for 3.x. We can discuss 4.x when Georgios is back.


Guillaume Smet

unread,
Aug 18, 2026, 7:38:42 AM (2 days ago) Aug 18
to Martin Kouba, quark...@googlegroups.com

I will merge it by the end of the day for inclusion in 3.39.

Georgios Andrianakis

unread,
2:12 AM (8 hours ago) 2:12 AM
to Quarkus Development mailing list, Martin Kouba
It's unfortunate there are so many different corner cases. 
I do believe it's safer to have this feature off in LTS

--
You received this message because you are subscribed to the Google Groups "Quarkus Development mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to quarkus-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/quarkus-dev/CALt0%2Bo-%3D7u%3DdjkHDvOGf4jKz2VQA-kKfyhvmWwXTRHC46jD-ZQ%40mail.gmail.com.
Reply all
Reply to author
Forward
0 new messages