added a disk bucket that stores new objects in the gaps (deleted fragments); 2) serialization of algorithm now uses a temporary file and only if successful, the specified file is overritten

* the LongStorage and LongAddress now has the ability to overwrite the stored object by its new version
* size of the new object must be exactly the same

 * the auto-close of disk storages now prints info to Log.INFO level
* LocatorStorageAlg now processes GetRandomObjectOperation

* Added netbeans files to ignore

* Added support for periodic check for temporary close

* Fixed generic type casting for the "named instances" map

* Added vector item access method for int-vector object
* Added backward-compatibility method for words-to-identifiers text conversion method
* Added support for var-arg constructors instantiation
* Added MESSIF test classes
* Added sample configuration files to testing resources

 * added an constructors to GetCandSetOperation

 * in Algorithm, now default is to print information about processed operations
 * MOPAW is now testing if the two object arrays are of the same size (in each .getDistance())

 * added a functionality to remove ".repeatEvery" thread from application

* Cleanup of parametric string conversion
* Added support to set operation logging verbosity in algorithm
* Added operationInfo config file action verbosity support

* Added simple extractor wrapper that allows to get data source from an URL that is provided in a parameter

* Changed classifier algorithm so that the parameters of the operation are used in classifier

   - specifically in BucketStorageExc and ExtractorException, at the moment
 * increased version to 2.1.6

 * modified behavior of '.repeatEvery' modifier of each action in CoreApplication - now the 'repeating thread' is removed after the loop ends (or if another .repeatEvery action with the same name is executed)
 * added the possibility to have RMIAlgorithm (client) with multiple threads:
   - the orig. RMI algorithm renamed to RMIAlgorithmOriginal
   - the actual connection to RMI server now encapsulated by RMIMethodExecutor
   - the new RMIAlgorithm manages a queue of RMIMethodExecutors, but by default it has only one connection (single thread)
   - new RMIAlgorithmMultiThread extends the RMIAlgorithm using multiple threads (and multiple connections) used in round robin manner
 * version updated to 2.8.5

* Updated classifier algorithm to work correctly with the classifiers that use operations as input and fixed correct conversion to distances
* Added support for creating named instance using multi-line parameters in config file

 * NavigationDirectory is not generixed any more 
 * version increased to 2.1.4

 * created an operation that encapsulates several (independent) k-NN operations (a batch of kNN operations)
 * the simple FileSeqScan can now process the batchKNNQuery in threads very efficiently

 * by given file with text represented data

 * modified field "runningOperation" of Algorithm so that the operations are not referenced from threads in a thread pool

 Created an operation (request) for refinement of a provided ranking operation by candidate set within the provided candidateSetOperation.

 * probably to be removed completely in future

 * ApproxKNNOp now properly works with its Arguments (and uses standard toString()) 
 * Algorithm processOPeration now puts an "output" parameter "OperationTime" to processed operation
 * MultipleOverlaysAlgorithm.setThreadPool now does not share the pool with the inner algorithms any more
 * fix in SortedCOllection - internal fields are now "volatile"

 * multiple-overlay algorithm now propagates the threadpool to the sub-algorithms, if they do not have any pool
 * minor version increased

* Updated pom file - new version and description

* Maven definition file added