<h3>SIG for Development of Persistence and Transaction Frameworks</h3>
<P>There will be a <a href="http://conferences.oreillynet.com/pub/w/15/bof.html"> Python Persistence BOF</A> at the O'Reilly Open Source Conference.
<P>Python Persistence <BR>Date: 07/25/2002 <BR>Time: 8:00pm - 10:00pm <BR>Location: Grande Ballroom C in the East Tower <BR>Moderated by: Patrick O'Brien, Orbtech
<P>A Python Persistence Special Interest Group was recently formed to explore ways to add basic persistence and transaction mechanisms into the core of Python to avoid duplication of effort by a variety of projects that have similar issues. This BOF will permit participants to ponder Python persistence in person. In addition, anyone interested in an informal Python Persistence breakfast discussion with Jim Fulton and Guido van Rossum is welcome to join us at the O'Reilly Food Tent Wednesday morning at 7am.
A number of projects are underway to provide persistence mechanisms for Python. These efforts have a number of common requirements, including:
<dl> <dt>Transparency <dd>Applications should not have to explicitly track object changes or
save objects. Applications should not have to explicitly query most objects. Typically, some "root" objects will be explicitly retrieved and other objects will be retrieved through normal Python object traversal. Application objects should not contain storage code, such as SQL statements or file operations, needed to provide their persistence. Application code may need to contain code to generate persistence-related events, but even these should be automated to the degree possible, for example by monitoring attribute accesses.
<dt>Transactional storage <dd>Data should be stored transactionally with support for rolling
back changes. Support for (optional) nested transactions should be anticipated.
<dt>Effective memory usage <dd>Applications should not use undue amounts of memory. Often,
persistence applications use databases that are too large to fit into memory. Objects should be loaded only when necessary and should be removed from memory when they are no longer needed.
- <p>Most of these efforts are focused on providing persistence using
- relational database data. The efforts are, for the most part, proceeding independently. Each will attempt to address the above requirements independently, with much duplication of effort.
- <p>The <a href="http://www.zope.org/Wikis/ZODB/FrontPage">Zope
- Object Database</a> (ZODB) has satisfied the above requirements for some time. ZODB is currently undergoing a transition from ZODB 3, which was based on ExtensionClass, to ZODB 4, which is based on Python 2.2 new-style classes. As a part of this effort, the ZODB persistence and transaction frameworks are being factored out of ZODB into separate packages, with the hope that they will be of use to other persistence-based frameworks.
- <p>It will be a huge duplication of effort if each of the various
- persistence projects has to address the above requirements independently. Worse, the resulting systems will have independent frameworks that are unlikely to interoperate. Objects built for one framework will need to be rewritten to work with another.
- <p>A new persistence-SIG is proposed to explore and, if possible,
- produce persistence and transaction frameworks that can be used for a variety of persistence implementations, including relational database-based persistence and the ZODB.
<p>Coordinator: Jeremy Hylton
- <p>Conclusion: When 1.0 versions of the frameworks are delivered, or
- September 1, 2003, whichever is sooner.
- <p>Deliverables: PEPs documenting the frameworks created and software
- implementing key parts of the frameworks.
- <p>Assuming that a satisfactory framework can be defined, then the
- framework and core implementations should be included in standard Python distributions.
The scope of this SIG includes common frameworks for:
<ul> <li>Transaction coordination
- <li>Basic persistence management, in particular observing object
- changes (to know when an object needs to be saved) and access (to know when objects are used so that unused objects are removed from memory), and
- <li>Activation and caching, to move objects into memory when needed
- and out of memory when no longer needed.
<p>The scope does not include:
<ul> <li>Concurrency control. This is the responsibility of specific data
managers that are plugged into the frameworks. The transaction manager simply tracks object changes and coordinates the activities of data managers to commit (or rollback) changes in an atomic fashion.
- <li>Query languages. Individual data managers or applications may
- provide query facilities. While it would be cool to have a common query facility for Python, and I would support such a project, that would be a different project than this one.
- <li>Integrity constraints. Some data managers, such as those based on
- relational data, will need to provide facilities for low-level integrity checks, while others, such as ZODB, will not. Similarly, garbage collection is the responsibility of the data manager, not the framework. Application-level integrity systems would also be interesting, but would not depend on a persistence system.