The Nutch 1.15 binary package and the Nutch job file will reach or even exceed 300 MB. A huge job file isn't ideal as it needs to be distributed in the Hadoop cluster. There are several reasons for this:
- missing exclusion of unneeded indirect dependencies, e.g.:
- indexer-elastic includes many Lucene libraries related to query processing
- huge plugins. That can be hardly avoided as these plugins provide also "huge" functionality, e.g., parse-tika is able to pass a long list of document formats. However, it should be possible to avoid overlaps between plugins (in case dependency upgrades can be made at the same time). Hot candidates are:
- any23 includes tika-parsers as does parse-tika
- lib-selenium and lib-htmlunit both use selenium and phantomjsdriver libs
Of course, users can easily build smaller packages by excluding unused plugins in src/plugin/build.xml. But would be good, at least, to try to reduce the size of the Nutch package.