View all newsletters
Receive our newsletter - data, insights and analysis delivered to you
  1. Technology
July 8, 1990


By CBR Staff Writer

As reported last week, (CI No 1,459), the big relational database guns have been rolled out, and their sights are firmly trained on the new start-up object-oriented database vendors. The battle plan has been drawn up in an extraordinary document called the Third Generation Data Base System Manifesto, written by Michael Stonebraker, developer of Ingres. The manifesto has been endorsed by Larry Rowe, Professor of Computer Science at Berkeley University and a fellow Ingres colleague; Bruce Lindsay, architect of IBM’s DB2; Jim Gray, author of Tandem’s NonStop SQL; Mike Carey (described as an object-oriented afficionado); Michael Brodie, director of database research for GTE; Phil Bernstein, director of DEC’s Cambridge, Massachusetts DBMS lab; and David Beech, technical advisor within Oracle. These senior statesmen of database management systems have decided that as the 1970s was the decade of the first generation, hierarchical database, and the 1980s were the years of relational database dominance, one thing the 1990s are not going to be is the era of the object-oriented database. Rather the third generation database, nameless at present, will be an extension of relational methods.

Persistent C++

The third generation database embraces three philosophical tenets. The first of these states that data management, object management and rule or knowledge management must all be addressed in the next generation of databases. In the data management stakes Stonebraker says that databases must be able to manage 100 transactions per second from 1,000 terminals; as for object management they must be able to store non-traditional data elements such as images and arrays; while in terms of rule management they must be able to apply rules about the data for integrity constraints and business processes. The relational heavyweights have decreed that systems that don’t address all three requirements will win only small niche markets – this is a reference to object-oriented databases, whose forte is object management, and which Stonebraker & co refer to as persistent C++, the vendors themselves being described as the O Companies. The second tenet of the Manifesto decrees that non-procedural access and data independence – both features of current relational databases – must be retained. Stonebraker refutes claims by the object-oriented vendors that computer-aided design users, which comprise the natural constituency for object oriented databases (CI No 1,449), do not need query language. He argued that managers of the designers using CAD applications want to run queries. The point here being that if the mighty relationals decree that SQL is necessary, object-oriented databases written in C++, will have to offer SQL, which slows their performance. The third tenet is interoperability: databases must interoperate with distributed database management systems, C programs, Fortran programs, business applications like Lotus 1-2 3, or software engineering tools. Therefore, database systems should not be bound too tightly with any specific language – are you paying attention persistent C++ vendors?

By Katy Ring

In addition to these three tenets,the manifesto offers 13 detailed propositions – the first collection of these refines object and rule management where the big relationals are most vulnerable: according to the manifesto, users can look forward to rich type systems moving beyond character string or integer types to x,y co-ordinate type points (so that a database will know, say, the geographical relation between addresses), and floating point number systems capable of dealing with an array as a data type (useful for records of employees whose salaries vary from month to month as the database could then manage an array of integers). Multiple inheritance, the core of object-oriented databases, will appear in relational systems so that a create statement can be extended to cope with inheriting data from previous fields arranged hierarchically. Functions will be installed on the database, accessed by SQL, enabling code to be shared among

applications so that whole records can be taken as an argument, not merely two integers. Rule management or, as the big relationals call it, referential integrity, will be installed on the database and will be universally applied by the database system. According to Stonebraker the successful database companies will be those that meet the propositions laid down in the manifesto and the race to get to the finishing line first is on. As might be expected he is tipping Ingres to win by 1993 or 1994, saying that Oracle and DB2 satisfy none of the manifesto’s propositions at the moment, nor do SQL2 or SQL3. He stated categorically that successful vendors will overrun the standards process because it is not aggressive enough. The vendors do have their work cut out, however, since such sweeping changes, implemented via proprietary SQL extensions, mean that they must, for example, rip out the hard coded type system in the heart of the database and rewrite it from scratch. The changes that these vendors make in their internal execution engine will lock object-oriented databases out from accessing relational third generation technology and will lock users more tightly in to vendor-specific databases and application design methods. Users, says Stonebraker, will have a hard time, and consultants will make a lot of money.


Content from our partners
Why food manufacturers must pursue greater visibility and agility
How to define an empowered chief data officer
Financial management can be onerous for CFOs, but new tech is helping lighten the load

However, with their sights so firmly trained on companies like Ontologic and Versant Technology, the relational guns have overlooked Interbase Software Corp’s database management system marketed by Cognos Inc as StarBase. For Cognos now has an application development environment (CI No 1,436) which already addresses the manifesto’s three tenets, and appears to be closer than any other vendor to meeting to the 13 propositions in the document. StareBase features triggers to enable business rules to be built into the database, BLOBS (Basic Large Objects) offering the storage of images or other unstructured data, two-phase commit support and a client-server implementation offering data placement anywhere in the network; by year end it will also store functions. Cognos has always maintanied that it was ahead of the competition – now its rivals are supplying it with documented evidence to support such claims. Qualms that Cognos has about its featherweight status or the fact taht StarBase is not an in-house product should be put aside and it should take the Third Generation DataBase Manifesto as the best marketing opportunity since Christmas. Please go for it, Cognos!

Websites in our network
Select and enter your corporate email address Tech Monitor's research, insight and analysis examines the frontiers of digital transformation to help tech leaders navigate the future. Our Changelog newsletter delivers our best work to your inbox every week.
  • CIO
  • CTO
  • CISO
  • CSO
  • CFO
  • CDO
  • CEO
  • Architect Founder
  • MD
  • Director
  • Manager
  • Other
Visit our privacy policy for more information about our services, how New Statesman Media Group may use, process and share your personal data, including information on your rights in respect of your personal data and how you can unsubscribe from future marketing communications. Our services are intended for corporate subscribers and you warrant that the email address submitted is your corporate email address.