OK, 3 and 4 done in r1080060 thanks! david jencks
On Mar 8, 2011, at 11:08 PM, Guillaume Nodet wrote: > I think for #4 it would make sense to use two interfaces. > > On Wed, Mar 9, 2011 at 01:58, David Jencks <[email protected]> wrote: >> I went ahead and committed this, let me know if there are any problems. It >> works fine for me so far.... >> >> I found the answer to (1) and (2) (feature event exports them) I think.... >> haven't had time to update for (3) and I'm still wondering about (4). >> >> thanks >> david jencks >> >> On Mar 4, 2011, at 5:02 PM, David Jencks wrote: >> >>> I spent a little time moving the jaxb tree for features.xml into features >>> core and getting it to work with features core. (and then a lot of time >>> trying to figure out how to get it onto my github branch. I think it's on >>> the "master" branch at https://github.com/djencks/karaf/branches) >>> >>> I have a few questions. >>> >>> 1. Why are the feature structure interfaces (Feature, BundleInfo, etc) >>> exported from feature core at all? >>> >>> 2. If they really need to be exported, is there a good reason to use >>> interfaces rather than the jaxb classes? >>> >>> 3. The schema allows 0..unbounded details elements since its an optional >>> member of a choice group. The original classes only allow one detail. I >>> guess we want to only allow one detail element? >>> >>> 4. There's only one Feature interface for both a complete feature (top >>> level in features element ) and a dependency feature inside a feature >>> element. The second one is more of a feature-ref since it doesn't have any >>> actual contents for the feature. I think it might be reasonable to have >>> two interfaces so as to distinguish these more easily. >>> >>> Does anyone want to review this or should I just go ahead and commit it? >>> >>> thanks >>> david jencks >>> >> >> > > > > -- > Cheers, > Guillaume Nodet > ------------------------ > Blog: http://gnodet.blogspot.com/ > ------------------------ > Open Source SOA > http://fusesource.com
