Change configure.classic to handle various versions of processors in a more sane way (this fixes several Solaris issues and atomics for i686).
Add fwdwait option to cms.request directive to allow pacing of forwarded requests (off by default).
Use native memory synchronization primitives when available when doing network I/O. This will eventually be extended to cover all other cases.
Add the qdl option to the cms.delay directive to allow changing the query window independently of the time a client is asked to wait for the query to actually complete.
Add 'pss.namelib' directive to allow proxies to pre-translate the lfn for servers that cannot do so (e.g., dCache xrootd door).
Optimize handling of shared-everything ile systems (e.g., dCache, GPFS, DFS, Lustre, etc.) in the cmsd.
Implement optional throttling for meta-manager requests in the cmsd.
New cmsd directive, cms.dfs, declares that the underlying file system is a shared-everything system (i.e., distributed file system) and allow for optimal configuration and meta-manager throttling.
Change the oss and fm components to use file extended attributes instead of meta-files. This affects copy, create, reloc, rename, and unlink in the oss layer. Migrate, purge, transfer, and most admin commands in the frm component. The 'all.export' directive now accepts the noxattr/xattr option. WARNING: If the migrator or purge options have been specified for any path in the 'all.export; directive then this change requires either the the 'oss.runmodeold' directive be added to the configuration file to provide backward compatibility or that the name and data spaces be migrated using the frm_admin command. See "Migrating tp Extended Attributes" manual for detailed information and the new 'frm_admin convert' subcommand.
Avoid physical copy if the operation can be handled using hard links. This greatly speeds up static space token reassignment.
Add platform independent interface to extended file attributes.
RPM packaging and Red Hat Enterprise Linux compatible init scripts capable of handling multiple instances of the xrootd daemons. The instances can be defined in the /etc/sysconfig/xrootd file and then handled using standard:
service xrootd start|stop|... service cmsd start|stop|... ...
or handled by name:
service xrootd start instance1 instance5
New '-s' commandline option for xrootd, cmsd, frm_purged and frm_xfrd creating a pidfile.
xrootd, cmsd, frm_purged and frm_xfrd now return failure to the shell when called with '-b' option (daemonization) and the daemon fails to initialize.
New 'EnableTCPKeepAlive' client environment option added enabling the TCP stack keep-alive functionality for the sockets. On Linux three addtional fine-tunning options are available:
New functionality handling process forking. When enabled (via the 'EnableForkHandlers' env option) prior to a call to fork it shuts down all the xrootd connection management facilities (including the connections themselves) and reinitializes them after the fork both in the parent and the child process. This ensures relative fork safety provided that all the XrdClient and XrdAdmin instances are closed when the fork function is invoked.