Log Search Slowness Analysis as a Fullsearch Primer
  ===================================================

  The files you'll need are:

    * src/lib/netdb/Log.java
      src/lib/netdb/Full_Search/Log_Search_Parameters.java
      src/lib/middleware/Full_Search.java
    * src/lib/middleware/NetDB_Connection.java

    with the starred files being optional, but included since they're passed through in the code path.
    I expect that the majority of the work you'll be interested will happen in Full_Search.java

  Flow:

    The client calls Log.full_search (...),

         which calls    Log_Search_Parameters.search (...),

         which calls    middleware.Full_Search.perform_search (...)

           to determine whether or not it's being run with a remote or direct connection.  RMI stuff
           happens (or not, if a direct connection is being used) and eventually the code path

         takes us to    NetDB_Connection.execute (...).

           This is where the call to generate the SQL is made, and subsequently executed.  The actual
           SQL generation

         happens in     Log_Search_Parameters.get_SQL (...),

         which calls    middleware.Full_Search.assmDisplayOrder (...)
                 and    middleware.Full_Search.assmCommonSQL (...)

           Each of the assm methods has a "Hashtable search_params" parameter. This Hashtable is built
           using the Log_Search_Parameters.legacy_search_params () method. It translates the parameters
           from the methods into the key value pairs that are currently hard coded in the legacy web
           front end. The hard coded values are processed in the assm methods to build the SQL (or part
           of it).

    It looks like the assm methods are where the majority of the complexity is. That's still Dmitri's
    code. I simply moved it into our classes. Fixing it (and all of full search) is the next thing on my
    list after the web front end.  I suspect that you'll be hard pressed to fix only log search, and not
    affect the other searches. If memory serves, they all use a very similar code path as discussed here.