Release date: 31 August, 2016
We are proud to announce the release of Fedora 4.6.0.
Resources
Team
Release Manager
- A. Soroka, the University of Virginia
Developers
A. Soroka, University of Virginia
- Aaron Birkland, Johns Hopkins University
Unknown User (acoburn), Amherst College
Aaron Elkiss, University of Michigan
- Andrew Woods, DuraSpace
- Benjamin Armintor, Columbia University
- Bethany Seeger, Amherst College
- Bill Branan, DuraSpace
- Esme Cowles, Princeton University
- Jared Whiklo, University of Manitoba
- Michael Durbin, University of Virginia
- Michael J. Giarlo, Stanford University
- Mirek Simek, University of Chemistry and Technology, Prague
- Mohamed Mohideen Abdul Rasheed, University of Maryland
- Nick Ruest, York University
- Peter Eichman, University of Maryland
- Yinlin Chen
Issue Reporters
Summary
The Fedora 4.6.0 release furthers several major objectives:
- Tighten the definition of the RESTful application programming interface (API)
- Refining the messaging service
- Use performance test fixtures to examine the effects of different backends
- Improve durability by encouraging the use of MySQL and PostgreSQL backends
- Fix bugs
This release will be the last release built against a version of Modeshape that uses Infinispan for data storage.
Changes
Messaging Interface
As the draft Fedora Messaging (SPI) specification moves toward finalization, the message serialization format has been modified to track the recommendations outlined in this document. This will affect any existing message consumers. There are four significant changes that messaging applications should be aware of:
- Previous versions of Fedora emitted messages with the header
org.fcrepo.jms.properties
. This header is no longer included in messages. - Previous versions of Fedora emitted messages with the header
org.fcrepo.jms.eventType
, using values with the http://fedora.info/definitions/v4/repository# namespace but which were not defined in the Fedora ontology. Theorg.fcrepo.jms.eventType
header now uses values from the newly published Event ontology: http://fedora.info/definitions/v4/event#. - A new header is included with the name
org.fcrepo.jms.resourceType
. This header includes allrdf:type
values of the resource in question. - Previous versions of Fedora emitted header-only messages where each header was prefixed with
org.fcrepo.jms
. The message serialization now also includes a body formatted in JSON-LD using the PROV namespace. All data found in the JMS headers are also available in the body serialization. Examples of this format can be found here: https://github.com/fcrepo4/fcrepo-event-ontology.
Please note: the JMS-centric header names are not part of the upcoming messaging specification and may be removed from a future version of the fcrepo-jms
module. Messaging clients are strongly encouraged to rely on data in the message body.
Application Programming Interface
One of the technical priorities of Fedora is to define a well-specified application programming interface (API) against which client applications can be written and future server-side implementations can be created. This Fedora API should be clear and detailed enough such that a corresponding technology compatibility kit (TCK) would be able to indicate if any Fedora implementation fulfills or diverges from the specification. With this in mind, several issues were addressed in this release that clean up Fedora's RESTful interaction.
An important particular change to note here is the move to use only weakly-validated ETags for RDF resources, in line with a correct interpretation of the HTTP specifications. This means that such ETags are no longer suitable for use with the If-Match
request header.
Web Access Control
Documentation of Fedora's implementation of Web Access Controls is available on the wiki.
Performance
The Performance and Scale group has been testing various versions of Fedora, including the 4.6.0 release candidates. Recent work has focused on running our JMeter test plans by multiple sites and testing the impact of using a relational database (MySQL or PostgreSQL) instead of LevelDB. The performance of the databases has been at least as good as LevelDB, and typically scales much better. In addition, testing has identified scalability issues with containers that link to a large number of other containers, and is working to address that issue. The current status of testing is tracked in the wiki: Performance and Scalability Test Plans.
Preservation
This release includes continuing support for various backend object stores. Default usage of that LevelDB backend has been removed, to encourage the intentional selection of a backend appropriate to the integration in hand.
Housekeeping and Bugfixes
Numerous refactorings, bugfixes, and clean-up tasks were addressed in this release: