response from the stager.
        Adjust the stalling of clients so that they are not delayed for more than 5 min.
        Use consistently the stagerhost variable set in the xrd.cf.manager file when sending
        requests to the stager.
        Fix the deletion of expired user requests - this part of the code should be rarely used
        given that we now have a proper way of cleaning up requests from diconnected clients.
        Move the definition of ReqElements for the .cpp file.
        Drop the createifnotexists opaque tag since it's not used anymore.
        Refector some error messages and fix formatting.

        Fix timing printout

The DriveCatalogueEntry class was initialising the host-name/server
to be used with the vdqm to an empty string and never changing it
there after.  This of course was causing the security features of
the vdqm to reject requests from tapeserverd to change its state.

Please note that the "integration" code of tapserverd is not fully
functional.  One should however be able to run a data transfer
session.

not tested as gridftp transfers are actually broken in SLC6 for long...

Conflicts:
	debian/changelog

Conflicts:
	castor/db/stager_oracle_create.sql

The heartbeat message from the data transfer session to the parent
tapeserverd process is broken.  Given the fact that the heartbeat
logic is not yet implemented in the parent tapeserverd process, I
am commenting out the sending of the message from TaskWatchDog to
parent tapeserverd process.

Mounting and unmounting a tape can take several minutes.

The different tape sessions, Cleaner, DataTransfer and Label
require TCP/IP proxies in order to request services from other
daemons such as rmcd.  Since the move of the sessions from
TapeDaemon to ProcessForker, the factories for creating those
proxies within TapeDaemon have become obsolete.

This implied adding file block ID tracking in TapeFile::WriteFile

PLEASE NOTE that the cleaner is not yet launched correctly.  I will
fix this in a future commit.

With this commit the TapeDaemon process no longer forks a session.
All sessions (CleanerSession, DataTransferSession and LabelSession)
have  been moved to the ProcessForker.

Properly use the now introduced custom exceptions in the case of transfers that failed to start because of the stager

and related to bug #105068: Draining looping when having shared pools and replicated files.

Renamed the new constraint introduced in DrainingErrors to respect the
convention. This is actually used in oracleGC.sql:483 in 2_1_14, not in
the master.
Fixed the constraint violation to be ignored by the GC.