Sorry JB I think you are wrong with this perception. A lot of the dev decisions do influence users, especially the ones were people build on top.
sent from mobile device Am 12.02.2016 12:53 nachm. schrieb "Jean-Baptiste Onofré" <j...@nanthrax.net>: > Hi David, > > For now, we are talking on the dev side of Karaf: from an user > perspective, it doesn't change anything (an user doesn't have to know bnd > to use Karaf). > > Anyway, projects using Karaf can still use the configuration they want: > maven-bundle-plugin without bnd files, bndtools, or whatever. > > Regards > JB > > On 02/12/2016 12:50 PM, David Daniel wrote: > >> I have had the same experience as Milen and Christian but I think I come >> to >> a different conclusion. I originally came to karaf because of the easy >> entrance into OSGI and we have a Maven build process that I could not move >> from. I think many other people come to karaf for those same reasons. >> Here is a conversation I had the other day where karaf was primarily being >> used for its build process and its tight integration of maven with >> features >> >> https://groups.google.com/a/saiku.meteorite.bi/forum/#!topic/dev/4uiWj1g2EU0 >> I think even bnd is getting away from having to put things inside of the >> bnd file and making sure you can configure osgi in other ways (Example >> would be making classes ending in impl private and using annotations to >> determine exports and imports through references. Even those elements you >> discussed previously such as runtime properties and system packages are >> more for the bndrun files so that bndtools can package and launch >> appropriately. I generally keep them out of the bnd file and let maven >> handle those elements for my project. If you want to maintain a bndrun >> file to test and use with bndtools that is different than requireing a bnd >> file. My general feeling is that almost nothing will need to go into the >> .bnd file soon (See this conversation >> https://groups.google.com/forum/#!topic/bndtools-users/yV_B5B1cU3k) so >> karaf should not require users to understand what it is. I think it will >> be optional soon enough so it is better not to require it now. On the >> other hand if developers like bnd tools then let them include an optional >> bndrun file that they are responsible for maintaining. I think with java >> 9 >> alot of people will come to OSGI for the tooling and it will be important >> to ease them into the OSGI build process as simply as possible. >> >> On Fri, Feb 12, 2016 at 6:28 AM, Christian Schneider < >> ch...@die-schneider.net> wrote: >> >> On 12.02.2016 11:46, Milen Dyankov wrote: >>> >>> For the record (in case you find some public evidence of me arguing the >>>> opposite) some time ago I was totally against using bnd.bnd files. After >>>> being kind of forced to do it for a while, I realized it doesn't really >>>> make any difference in the effort that it takes to maintain those and at >>>> the same time provides clearer separation of concerns. So I changed my >>>> mind >>>> :) ! >>>> >>>> Interestingly I have the same experience. I first hit bnd.bnd files in >>> some pax projects and thought they were a strange way to configure the >>> OSGi >>> setup. So at this time I was also rather against it. The more I worked on >>> it in pax and also while I did some experiments in bndtools the more I >>> liked the >>> style of bnd files. At some point then I migrated the Aries JPA project >>> to >>> it and I really like the result. >>> >>> Christian >>> >>> >>> -- >>> Christian Schneider >>> http://www.liquid-reality.de >>> >>> Open Source Architect >>> http://www.talend.com >>> >>> >>> >> > -- > Jean-Baptiste Onofré > jbono...@apache.org > http://blog.nanthrax.net > Talend - http://www.talend.com >