On Wed, Aug 17, 2011 at 08:53, Clint Popetz <clint(a)42lines.net> wrote:
I have a proposal from one of the wicket committers (Igor Vaynberg, who is
also one of my employees) to transition seam-wicket to become a wicket
module that integrates wicket with weld, so that it's supported by the
wicket team. This is a net win, in my opinion, because (a) the only code in
seam wicket is really just code to configure wicket's request cycle to
start/stop conversations and perform injection and has no other seam
dependencies, (b) this allows the release to be correctly synced to the
wicket releases, which we currently lag and are thus not compatible with,
and (c) he has more time to maintain this than I do, and would do a better
job of it.
Is this acceptable to the seam team? The only thing I really need is for
the weld 1.1.1 artifacts to be in the central m2 repo, because wicket is
published there and the central repo doesn't let you have dependencies on
non-central-repo artifacts. Is that reasonable/possible?
In general, I like the prospect of the Wicket integration being supported by
the Wicket team. That's definitely a step forward.
However, I think integrating directly with "weld" is a step backwards (at
least by putting weld in the name, it throws in the towel). The recommended
approach for CDI support is to integrate with the CDI APIs and use SPIs to
wrap integrations that are dependent on the implementation. From my
understanding, those SPIs will likely go away anyway in CDI 1.1 once the
standalone bootstrap API is in place.
In summary, +1 for the module to be supported by the Wicket team, -1 for
using "weld" in the name (and throwing in the towel on CDI portability).
I would still like for the Wicket integration to feel part of the Seam
community. We have other CDI integrations that follow this model where they
are bundled w/ Seam and used in Seam examples, but maintained on the side of
the integrated technology.
Principal Software Engineer, Red Hat | Author of Seam in Action
Registered Linux User #231597