Hi,

Sorry, but you do not have to delete the workspace directory. Not sure where you are getting this from? In cases where you have a large number of packages/assets in Guvnor it is rather recommended _not_ to delete the search indexes written onto the file system, because it takes extra time to re-create them.
 
May be it is recommended. But when you modify a lot the assets, rename, copy, delete,etc.. Guvnor gets lost.
 
At startup, Jackrabbit (containent in Guvnor) reads all the database and creates the two directories. You do not need to backup them.
Again, I don't know where you are getting this from. Jackrabbit does _not_ read the entire DB on startup and does not write any of the JCR content stored in an RDBMS to the file system if you have configured it to store to the RDBMS. If you for any weird reason are seeing this on your end, you should really have another look at your repository.xml.


The search indexes, they are stored on the file system no ? And with no index, you cannot reach any of the assets. So when you delete thoses directories, at startup, they are re-created and the database is read. I do not know what is read, but it can take quite some times, depending on the size of the package. Now If there is a way to store thoses indexes in the database, I would be happy to know how.

The project I am on has many rules and many big web decision tables that end up with more than 100000 rules.  I am using 5.3 snapshot with MVEL beta6 => Startup time =5 minutes and building the package, 3 to 5 minutes on a HP G6 processor. On my PC that has an i5 processor and a 32 bits linux, I cannot build anymore the package. 
With those sizes, which is not that much, Guvnor/jackrabbits gets lots on the indexes : it gives jackrabbit exceptioorn or spaces are not considered. So then, what I do is delete those two directories and restart guvnor and everything is fine again.
Thanks
Nicolas