I'm not really interested in continuing this conversation, but to summaries what I've said before, and give pointers to a couple of the places I think would have interesting followup.
The tooling for making an OSGi app is much better than the last discussion I was involved in[1]. Also, Tika is now OSGi complaint[2] (which would make life tons easier), than the last time there was a discussion. The big trick IMHO, is still how to get the underlying Hadoop to be OSGi aware or friendly, and package everything into a job, and get Hadoop to unpack it in such a way that the OSGi implementation will work correctly. The plugin model Nutch uses is very similar to how OSGi works, so it should be fairly straightforward to port. I had a version of Nutch that semi-worked based upon a 1.0 or 1.1 branch. The big problem was how many things would needed to be re-worked, while there was a pretty significant re-write/re-org in progress. I wouldn't bother with making Nutch OSGi friendly, until Hadoop can distributed jobs as OSGi bundles. It should be possible, but I haven't wanted to take the time to dig into it. But there are guys on the Hadoop list who have made Hadoop run as inside an OSGi framework [3]. Not sure if they distributed jobs as OSGi bundles or what. I've never taken the time to investigate them. It just doesn't seem to be high on the list of priorities of the Hadoop folks to get that done. The problem with OSGi is that so many third party libraries assume global classloaders to implement their plugin systems. Nutch used to be plagued by code which didn't have good release controls (at one point it was using at least one version of an Apache project that was never released, and had been abandoned in place, commons-console I think? It also wasn't obvious the precise versions of several other libraries, I think that the usage of Ivy has addressed all of those issues). I think it'd be great to see this done, but I never had the time to actually make it happen. Hopefully the pointers get you to the proper discussions and points where similar and related work are being done. Kirby [1] http://www.mail-archive.com/[email protected]/msg02694.html [2] http://tika.apache.org/0.6/index.html [3] http://mail-archives.apache.org/mod_mbox/hadoop-common-dev/201202.mbox/%[email protected]%3E On Wed, May 30, 2012 at 6:45 AM, Lewis John Mcgibbney < [email protected]> wrote: > please see the countless conversations we've had on this on both user@ > and dev@ list archives. > > Thank you > > > > On Wed, May 30, 2012 at 5:36 AM, blunderboy <[email protected]> > wrote: > > Can we run Apache Nutch 1.4 in OSGI framework. I want to create an OSGI > > bundle of nutch. I am using eclipse indigo to compile the Nutch source > code. > > So i think there should be some kind of plugin which can create OSGI > bundle > > instead of jar after the compilation. I just need the OSGI bundle of > Nutch. > > > > I don't even know if it is possible. > > > > -- > > View this message in context: > http://lucene.472066.n3.nabble.com/OSGI-bundle-of-nutch-tp3986767.html > > Sent from the Nutch - User mailing list archive at Nabble.com. > > > > -- > Lewis >
