i've pushed the preview to [1]. if there are no objections, i'll commit it by the end of the week.
regards, gerhard [1] http://s.apache.org/kIn 2013/8/16 Gerhard Petracek <[email protected]> > hi luciano, > > i haven't checked it across the whole code-base. > > @ tuscany: > great, but as you see it's hard to know which parts of the > internal-packages need to be stable, because they are used by other > projects like tuscany... > > @ "specification" vs "wink specific": > imo it's clear that e.g. org.apache.wink.server.spi contains the spi of > wink and not a spec. spi (independent of the content). > > @ osgi: > yes - osgi is always special. as far as i was told osgi had issues with > java.util.ServiceLoader. if that's (still) true, it's at least possible to > fix it. > > i'll provide a preview for LifecycleManager. based on that we can discuss > further topics (e.g. a spi which allows apache tomee to use wink as jax-rs > implementation). > > regards, > gerhard > > > > 2013/8/15 Luciano Resende <[email protected]> > >> Is the current differentiation used to identify wink pieces that are part >> of the spec versus things that are "proprietary" to Wink ? Other then >> that, >> I don't think we are doing any kind of enforcement on the usage of those >> internal methods (e.g. OSGI visibility or something), and I have used >> plenty of those when doing the deep integration with Tuscany. >> >> But it should be ok to have a better extensibility mechanism, but then we >> might have to identify what is "specification" versus "wink specific". >> >> >> On Wed, Aug 14, 2013 at 2:39 PM, Gerhard Petracek < >> [email protected]> wrote: >> >> > hi luciano, >> > >> > e.g. WinkConfiguration, DeploymentConfiguration, LifecycleManager and >> many >> > others are in *.internal.* packages. >> > technically it's possible to extend/re-use them, but they would violate >> a >> > clean spi (interface/s in a spi package would import internal class/es). >> > >> > a first step would be e.g. to support custom LifecycleManager/s which >> can >> > be implemented without internal classes. >> > >> > regards, >> > gerhard >> > >> > >> > >> > 2013/8/14 Luciano Resende <[email protected]> >> > >> > > On Wed, Aug 14, 2013 at 12:51 PM, Gerhard Petracek < >> > > [email protected]> wrote: >> > > >> > > > hi @ all, >> > > > >> > > > i was going to start with WINK-397. >> > > > however, just adding new interfaces and using them via >> > > > java.util.ServiceLoader isn't enough. >> > > > a lot of central classes (as well as interfaces) are in one of the >> > > > *.internal.* packages. >> > > > -> to get a clean spi, we have to move some parts. >> > > > if there are no objections, i'll create a first draft based on the >> > wink2 >> > > > branch. >> > > > >> > > > regards, >> > > > gerhard >> > > > >> > > >> > > >> > > Could you clarify a little more about what issues you are seeing and >> what >> > > changes you are planning ? >> > > >> > > -- >> > > Luciano Resende >> > > http://people.apache.org/~lresende >> > > http://twitter.com/lresende1975 >> > > http://lresende.blogspot.com/ >> > > >> > >> >> >> >> -- >> Luciano Resende >> http://people.apache.org/~lresende >> http://twitter.com/lresende1975 >> http://lresende.blogspot.com/ >> > >
