[This article appeared in the October 1994 issue of the Cache Update newsletter]


                     AFS 3.4 New Feature Summary
			     25-Aug-1994


This is an outline of the new features to be included in AFS 3.4.  The
list is not all-inclusive and doesn't address any defect fixes, but
instead provides an overall summary of the upcoming enhancements.
Complete details of these new features will be made available in the
AFS 3.4 Release Notes documentation that will accompany the final
release.

The primary focus of the AFS 3.4 development effort has been to
improve the overall performance of AFS.  Major changes were made to
streamline the cache manager and file server.  Performance
improvements were also made to the backup system and several other key
AFS components.  The majority of these changes don't involve user
interface changes, so won't be described in this article.

The backup system also received several functional improvements to
better support customers running their backups in an automated
fashion.  To support tape stackers and jukeboxes, external programs
can be run just before mounting and dismounting tape media.  Customers
can create scripts that send tape handling instructions to their
automatic tape handling hardware, for example.  Improved control over
user messages and prompts have also been added.  Both features are
enabled through a new per-device configuration file.  Finally, the
"-localauth" and "-cell" command line options have been added to
"butc" and "backup", making it easier to provide authentication for
these commands.

In the past, AFS did not behave consistently on hosts having more than
one network interface.  Using these multi-homed machines as AFS
clients or servers caused problems with the callback system.  In AFS
3.4, we introduce partial support for this configuration.  Simple file
servers may be multi-homed, though database servers and clients are
still not supported.  The volume location database no longer records
the location of volumes using the file server's IP address, but
instead uses a unique identifier assigned to the file server.  This
provides a consistent reference to a file server no matter which
network interface is used for AFS traffic.

AFS 3.2 introduced a "preferences" feature that allows a cache manager
to select which file server is chosen when multiple "read-onlies" are
available for a given volume.  A similar preferences feature has been
added in AFS 3.4 to bias the cache manager's selection of database
servers.  This feature will aid those customers who have database
servers spread out across a wide-area network since it will allow
clients to primarily communicate with the closest server instead of
choosing one at random.

Smaller changes in AFS 3.4 include the "vos create -quota" option to
allow quota to be set on new volumes without mounting them first.  The
implicit access granted to members of "system:administrators" has been
increased from "a" to "la".  The number of AFS partitions supported
per file server has been increased from 26 (vicepa - vicepz) to 256
(vicepa - vicepiv).