Re: [krazo-dev] Discussion about future of krazo-cxf

[Christian Kaltepoth <christian@xxxxxxxxxxxx>]

Hi Tobias,

thanks for starting this discussion.

I agree that CXF support isn't a high priority anymore. Especially, because OpenLiberty started to use RESTEasy instead of CXF. 

And I also spent quite some time in the past debugging CXF issues. Unfortunately the issues I created in the CXF issue tracker are not addressed even after about 2 years (see CXF-7834 for example). And to be honest, I don't think that I want to spend even more time on it.

But I'm not completely sure if we should completely remove this module. Of course currently it is just experimental and not useable for real work application, but maybe somebody with more CXF experience will pick it up later on and help us to improve it. However, we should definitely make clear, that this module is experimental and not supported in any way.

Other thoughts?

Christian

Am Sa., 3. Apr. 2021 um 20:08 Uhr schrieb Tobias Erdle <tobi.erdle@xxxxxxxxx>:

Hi all,

with this mail I'd like to open a discussion about the future of the
Krazo CXF integration. We talked a few minutes about that topic during
the MVC call, but decided that this is more a topic for the Krazo
mailing list.

In my opinion, we should think about dropping the CXF support in the
next major release and deprecate it in a minor release. My reasons for
this are:

- CXF seems to getting replaced in OpenLiberty by RESTEasy during the
update to Jakarta RESTful Web Services 3.0 (see
https://github.com/OpenLiberty/open-liberty/issues/11803). That means,
that the big application servers all use JAX-RS implementations we are
supporting really good and only TomEE is left with CXF, which is, in my
perception, a niche product.
- Most of our workarounds are CXF related and investigating issues costs
a lot of time because of little documentation and a really big codebase
of CXF itself.
- I tried to understand CXF Bean Validation to fix an Krazo issue and
asked for help on the mailinglist, but didn't receive a single anwser. I
don't know if it was only bad luck or if they aren't really interested
in this topic. Anyway, I think it's hard to implement and maintain an
integration for a framework where is only little information available
for deep technical questions.


So that are my two cents onto this topic. What are you thinking about this?


Best regards,

Tobias

_______________________________________________
krazo-dev mailing list
krazo-dev@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/krazo-dev

--

Christian Kaltepoth

Blog: http://blog.kaltepoth.de/

Twitter: http://twitter.com/chkal

GitHub: https://github.com/chkal