On Mon, Sep 13, 2010 at 8:54 AM, Thomas Müller <[email protected]>wrote:
> Hi, > > I think I understand your use case, but I'm afraid all I can tell is: > it won't be easy. With the current Jackrabbit implementation, you need > both a PersistenceManager and a FileSystem implementation. To run > queries, you additionally need to use a SearchIndex implementation, > the default one uses Apache Lucene (which in turn is not using the > Jackrabbit FileSystem interface). Also, for large files most likely > you want a DataStore implementation (probably the easiest to > implement). Clustering uses it's own storage component, but maybe it's > easier in your case to ignore that and write your own clustering, > because the Jackrabbit clustering is based on the assumption that the > PersistenceManager is shared for all cluster nodes. > > yeesh. Well, one of the things that my datastore mechanism provides is cluster-neutral clustering - you don't have to have the different pieces aware of each other to be able to cluster sanely. We can handle Lucene, too, but you're right - the multiple storage mediums turn into a hassle. (I was going to cross that bridge when I got to it.) > Implementing all those components is a lot of work, and all that would > be lost once Jackrabbit 3 is ready. > > Well, we can only hope that Jackrabbit 3 hurries up, I suppose. -- Joseph B. Ottinger http://enigmastation.com
