1. A Generalized Modeling Framework for Schema Versioning Support
Fabio Grandi, Federica Mandreoli, Maria Rita Scalas
CSITE-CNR, DEIS
University of Bologna, Italy
The title of my talk is ODMG language extensions for generalized schema versioning support, a work which has been done by myself and Fabio Grandi.

2. A generalized OODB Model with Schema Versions
Support of consolidated and temporal schema versions
Support of a complete set of primitive schema changes
Provided operational semantics
My talk will be articulated in the following points:
First I will present a brief overview of the research project to which the present work is one of the contributions,
then I will introduce the proposed extension the the ODMG languages for generalized schema versioning support.
Finally, I will present conclusions and future work.
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

3. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
The Research Project
Extension of the ODMG standard
for schema versioning support (branching approach + temporal approach)
at level of
Object Model ( OODM|SV+ )
Low-level operations (schema changes)
High-level language (ODL+OQL)
This work has been developed in the framework of a research project whose aim is the extension of the ODMG standard in order to support schema versioning in Object Oriented Databases. The generalized schema versioning we consider includes both temporal schema versioning, as adopted in temporal applications, and branching versioning, usually adopted in design applications.
This extension has been partially done at the Object Model level and the outcome is OODM|SV+, which is a model which integrates the pure temporal schema versioning with the branching approach,
whereas this paper present the concerned extensions at O D & Q Language level.
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

4. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
Schema Versioning
Allows information systems to “survive” after schema re-organizations (schema evolution)
Allows legacy applications to continue to work after schema changes are effected
Allows to access/modify any available data through any available schema (full schema versioning)
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

5. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
The Branching Approach
Early requirements for versioning arose in advanced application domains like CAD/CAM and software engineering
Versioning introduced to support different users working on parallel, or even merging, versions of the same piece of data
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

6. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
Branching Schema Versions
Schema versioning involves intensional (structural) aspects of data objects
Goal: allow a “safe” management of
different schemas used at the same time
Schema versions organized as a DAG (consolidated / design versions)
SV1
SV3
SV4
SV2
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

7. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
The Temporal Approach
Versions have recently shown useful also for other object-oriented applications:
GIS
Multimedia applications
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

8. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
Temporal Schema Versions
Schema versions have a temporal pertinence
Allows the management of the history of the structure of data objects
Allows the system to keep track of the history of effected schema changes
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

9. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
Temporal Schema Versions
Provides a complete support to extant data access by means
of past, present and future
schema versions
Time line
SV1
SV2
SV3
SV4 t0 t1 t2 t3
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

10. A New Integrated Approach
A generalized framework for the support of design and temporal schema versions
Aimed at improving the application potentialities:
full control of the design process in engineering applications
concurrent scenario analysis in temporal databases (bifurcation of the time line)
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

11. The Underlying Model: OODM|SV+
A database is a set of versions organized as a DAG
Each version has a version coordinate set
a bitemporal timestamp
a user defined name (label)
Versions with the same label belong to the same database node
Derivation relationships between nodes are defined by edges
In OODM|SV+ a database can be seen as a set of versions organized as a DAG.
Each version is made up of a schema version and the corresponding data pool and is referenced thought a set of version coordinates that is a
a bitemporal timestamp for temporal schema verisoning support and/or
a user defined name, named label, for branching versioning support.
All versions with the same label share a common property and, for this reason, belong to the same node.
Derivation relationships between nodes are defined with edges in the DAG.
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

12. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
Basic Components
Time
represented as: TIME={0,1,…,now,…,}
bitemporal domain: TIMEt TIMEv
Temporal version: TV=(SV,SDV) representing a homogeneous state of the world with a unique time pertinence
Schema Version (SV): intensional component
Stored Data Version (SDV): a consistent instance set
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

13. F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000
The Database Concept
A database is a function DB: (L  TIMEt TIMEv) TVS {}
It associates one temporal version to each time-point and label vt
TV1
TV2 tt L1 vt vt
TV1
TV1 tt tt vt tt
TV2
TV1 L4 L3 L2
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

14. Schema Change Primitives
The set of schema change primitives
allows non-destructive changes to consolidated versions
allows the generation of a new consolidated version (DAG node)
allows the integration of features of different consolidated versions
allows the addition of a new semantic derivation (DAG edge)
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

15. The Taxonomy of Primitives
Modifications acting on nodes:
on attributes, relationships, operations and exceptions
on hierarchical relationships
on the type collection
Modifications acting on edges:
pick-, new- and merge-type
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

16. Semantics of Primitives (1)
Starting from:
a schema change u
two version identifiers ss1, ss2L TIMEv
a schema change validity sc
ss1
ss2
TV1
TV2 L1 L1 u TV sc L1
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

17. Semantics of Primitives (2)
The TVU function:
generates a new TV by applying u to the selected TVi
The BDBU function:
selects the TVi identified by ssi
places the new TV in the database with the assigned validity sc
ss1
ss2
TV1
TV2 L1 L1 u TV sc L1
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000

18. Conclusions and Future Work
we proposed a new generalized model for the management of schema versions in OODBMS with a complete set of schema changes
Future Work
study of semantic and implementation issues
addition of other versioning dimensions (e.g. space)
We have proposed and ODL/OQL extension for generalized versioning support.
Our future work will be devoted to
the definition of a detailed semantics for the proposed language extensions based on the underlying model
the addition of other versioning dimensions like spatial ones.
F. Grandi, F. Mandreoli, M.R. Scalas - ADC 2000