DATABASE STRATEGY
A BORLAND WHITE PAPER

B O R L A N D
1800 Green Hills Road, Scotts Valley, CA 95067, Tel. (408) 438-8400



Table of Contents

Section I
Overview
Section II
Product Directions
*Operating System Platforms
*Interbase
*dBASE
*Other Products
Section III
Next Generation Database Architecture
*Developing Applications Using OOP
*Common Development Tools
*User Interface
*The Borland Database Engine
Section IV
Interbase
*Current Interbase Architecture
*Future Interbase Architecture Enhancements
Section V
Summary
*From the Leader in Customer Satisfaction
*A Winning Strategy



Overview

This paper discusses Borland's database strategy as it relates to the
Paradox, dBASE, and Interbase product lines. Driving this strategy is
Borland's commitment to provide customers with industry-leading technology,
service, and support through its database products.

The foundation for this strategy is to provide corporations with the
leading-edge technology to overcome the computing challenges of the 1990s.

During this period corporations will need to address several key trends in
computing:
*Emerging Graphical User Interfaces. The movement to graphical based
applications is occurring quickly. This trend is expected to continue as
software manufacturers invest the lion share of their R&D into the
development of graphical products.
*Object Oriented Methodologies. Just as the introduction of the
integrated circuit revolutionized computer hardware, so also the introduction
of object oriented methodologies is revolutionizing the way in which software
products are built and operated. Borland's commitment to object orientation
will allow for faster development of new products and versions, the ability
to reuse development work across a variety of products, increased reliability
of released products, and greater consistency and ease of use for users.
*Client/Server. Enhanced connectivity to corporate data is a key
requirement for companies looking to remain competitive in the 1990s.
Borland is committed to supporting both SQL access to remote databases, as
well as providing access to a broader variety of data in ways that will
deliver optimal performance. This will include support for distributed
database access and support for objects and complex data types such as
graphics, sound, and other binary objects.
*32 Bit Computing. Over the next year we will see the emergence of true
32 bit operating environments on the PC platform. The new systems will
provide a variety of benefits to users and developers, including support for
high-performance and true multi-tasking environments.

In addition to these trends, corporations must deal with many other changes
over the next few years. New operating systems will emerge that offer a
variety of capabilities, but require many decisions. Changing hardware
architectures, such as RISC processors, will alter the purchase patterns and
decisions facing corporations. Connectivity issues are becoming a major
challenge as corporations attempt to connect all of their hardware platforms.
Downsizing of applications and distribution of data to a broader set of end
users present as many challenges for corporations as opportunities.

Borland 's objective is to partner with corporate customers to provide the
technological leadership, tools, and support to overcome these challenges.
The following pages outline Borland's database product directions and
strategies in order to deliver these solutions to customers.



Product Directions

"By bringing two database platforms under one strong management and support
organization, Borland is providing a unified solution to our current database
needs."
- Carl Hane, Group Leader Micro Applications Systems, MCI International, Inc.

Operating System Platforms

With the Ashton-Tate merger, Borland can provide its customers with a suite
of application and language products across a variety of operating system
platforms.

Borland product line and platform coverage

Interbase

The Interbase server is the first relational database management system to
satisfy the demands of on-line complex processing applications used in
finance, banking, and manufacturing. Designed for optimal performance in
multi-platform networks, Interbase provides features targeted at the needs of
highly complex database applications in dynamic, distributed environments.

Interbase is clearly the technology leader in the database server market,
with advances such as a true peer-to-peer architecture, highly efficient
event alerts, smart automatic two-phase commit and rollback, multi-threaded
servers and instant recovery from hardware failures. In the future,
Interbase will be tightly linked to Borland's development tools and
languages, including dBASE and Paradox.

Interbase 3.0, the current version, supports comprehensive distributed
database processing among a wide variety of industry leading platforms,

including the entire HP-UX line, HP's entire Apollo line, VAX/VMS and
VAX/Ultrix lines, the DECstation and all Sun workstations. This connectivity
will be expanded to a variety of other platforms including 32 bit systems,
like OS/2.

dBASE

"Now we know that dBASE will have a bright future in Borland's hands."
- Adam Greene

Borland will support the dBASE IV cross-platform strategy, by continuing to
build additional dBASE ports. Borland believes that this cross-platform
strategy is what customers want, and that it encourages them to standardize
on dBASE.

Other Products

The majority of Borland's products are comprised of applications and tools to
meet the challenges of today's corporations. This includes Borland language,
database, and spreadsheet products, and also includes future database
compiler products. These applications are provided on the DOS platform and
will be made available on other personal computer operating system platforms
as the corporate challenges necessitate. For today, the next platform will
include Windows. For tomorrow, these platforms could include OS/2, Windows
NT, UNIX, or other 32 bit operating systems.

DOS Platform

Borland will continue to support and enhance both dBASE and Paradox in the
DOS environment. Borland believes that the DOS operating environment will
remain strong for the immediate future. Even as the graphical environments
and 32 bit operating systems gain greater acceptance, Borland believes that
continued support of the DOS platform is critical to corporations.

Paradox

Paradox for DOS is the award winning relational database management system
that allows you to ask complex, ad-hoc questions about your data without the
need for programming. The Paradox Application Language, or PAL, is a
powerful, full-featured development language for programmers to develop
powerful custom applications.

Paradox can also be used as a front-end to remote SQL database servers. With
Borland's SQL Link product installed with Paradox, a Paradox user can easily
and transparently access remote SQL data using familiar Paradox commands.
Servers currently supported in SQL Link are Microsoft/SYBASE SQL Server,
Oracle Server, IBM EE DBM, DEC Rdb/VMS, and by first quarter 1992, Novell SQL
and IBM DB2 through the Micro Decisionware gateway.

dBASE

Borland is committed to revitalizing dBASE, delivering more power, more
flexibility, and better support to the entire dBASE community. Borland is
committed to releasing major new DOS versions of dBASE to protect customer's
investments on this platform.

In addition, Borland intends to use its object-oriented language technology
to accelerate releases of new versions and introduce powerful new features,
including new SQL database server links, event handling, enhanced
performance, and object-oriented extensions to the dBASE language.

In 1992, customers can expect to see a new dBASE for DOS. Customers can also
expect to see enhancements to the developer products, including a compiler
that will generate fast, tight code. The compiler will generate DOS .EXEs
that can run stand-alone without a runtime.

Borland will also add new SQL database server links to dBASE including DB2,
Oracle, Interbase, and others.

Graphical Operating System Platforms

Borland plans to introduce a new generation of products that will ease
customer transitions into the world of graphical user interfaces for all
Borland's leading applications and programming languages. Borland's
graphical products will provide the benefits of an object oriented user
interface (OOUI). These products will first be available for Microsoft
Windows 3.0. Future products for OS/2 and other 32 bit operating systems
will provide a tremendous increase in computing power, and Borland will lead
the way with applications that take advantage of that power. Borland is
evaluating early development phases of products on these other platforms as
new operating systems move into the marketplace.

"Borland is once again going to revive a marketplace that was stagnating; the
users will win."
- Jeffrey Tartar
SOFTletter

Paradox for Windows

Paradox for Windows is the first version of Paradox to take advantage of
Borland's new object-oriented architecture. Objects are easily managed and
manipulated using a "SpeedBar" to build database applications for all levels
of users. Also, they have a powerful, interactive, context-sensitive "Menus
On Demand" capability that allows users to easily access the Paradox for
Windows database power.

Borland's Paradox for Windows will access data wherever it is, fast. It can
directly read and write multiple database file formats, such as Paradox data
files, dBASE, Novell Btrieve, spreadsheet data, and a variety of SQL database
servers. Paradox for Windows can also support many data types including memo
fields, graphics, binary types, and more.

Paradox for Windows is ideal for developers as well as end users. As the
first database with object-oriented programming techniques, not only can
Paradox offer more powerful applications, but the time necessary to develop
them is significantly reduced.

dBASE for Windows

dBASE for Windows will share the same object technology for creating forms,
reports, and other objects as Paradox for Windows. Additionally, dBASE for
Windows will transparently access external data files.

The difference between the two products is the programming language.
Borland's new dBASE for Windows will be dBASE III+ and IV language compatible
and will be able to run DOS-based applications in the same manner as Paradox
for Windows runs DOS-based PAL applications. Furthermore, the dBASE language
is being extended to support Borland's object oriented programming graphical
user interface and the Borland Database Engine functionality.

In addition dBASE for Windows will also have a dBASE language compiler. The
compiler will leverage Borland's world-class compiler expertise, and produce
tight, fast .EXEs for either the DOS or the Windows environment. Borland's
Integrated Development Environment (IDE) as found in Borland C++ and Turbo
Pascal for Windows will be incorporated into dBASE for Windows giving dBASE
developers access to the most powerful development environment available. In
addition, Borland's Resource Workshop which allows developers to easily
create and manage dialogue boxes, menus, icons, and other resources, will
also be included.



Next Generation Database Architecture

As represented first for the Windows operating system platform, Borland is
developing a family of products that take advantage of the graphical
environment. This section describes more fully Borland's next generation
database product architecture.

Developing Applications Using OOP

Borland's object oriented programming (OOP) technology provides tremendous
flexibility and power in building applications for its customers. Borland
uses C++ for its development, which produces code that is easily
maintainable, extensible, and reusable. That means Borland can develop
software far more quickly, and share functionality between products more
easily, by re-using code modules (objects). The resulting programs are also
easier to test, and so they are more reliable when released.

Borland's object-oriented approach supports the rapid development of core
technologies that are shared between products. For example, shared
technology can be seen across Borland's language spreadsheet and database
products.

Common Development Tools

Under the DOS platform, a decision to program database applications in dBASE
or Paradox's PAL currently carries several implications. For example:
*what kind of data can be read and written
*what level of control is given to the developer
*how much time is needed for development.

Under graphical platforms, using Borland's object technology, Borland
believes that choosing a database language should be based on the language
choice; such as between traditional programming languages, C++ or Pascal.
Borland's language products include the same debugger, profiler, linker, and
Integrated Development Environment (IDE). Therefore, the decision to develop
a database application in dBASE or PAL will also become a decision based on
which language is most appropriate for (or familiar to) the developer.

User Interface

Borland plans to introduce a new generation of products that will ease
customer transition into the world of graphical user interfaces. Borland's
graphical products will be based on user management and manipulation of
objects. This usage of objects as the starting point creates what we call
object oriented user interface (OOUI).

What is an Object?

Just like in OOP, user interface objects consist of a set of properties and
methods that perform actions. For example, if you have a button object, the
button may have location properties, a title, color, and also methods that
define actions based on an event (e.g., an action when the button is pushed).
So, encapsulating property and methods together is an important part of the
database interface object model.

Therefore, properties are the fundamental components of an object. Different
objects have different properties. Database components, such as tables,
forms, and reports are made up of objects (the grid, the data, etc.) each of
which have its own set of properties.

Objects are a collection of properties and methods.

The user interface in Borland's applications uses objects; in fact, the user
has a range of different objects from which to choose. Everything you see
on-screen is an object. Whenever you want to change or inspect an object's
properties, the user right-clicks on the object. This displays a Menus On
Demand (menu of the object's properties and methods) unique to the object
type. Selecting an object property from the menu choices allows the user to
customize the object directly.

An object is self-contained, meaning that all the actions it can perform and
all the ways in which it can be changed are part of the object itself.
Because objects are self-contained, you go to the object directly when you
want to do something with it or change something about it.

Object Method Languages

One of the properties of an object is a set of methods, or actions which are
based on a predefined user response. Borland uses a procedure called, "late
binding of methods" to object properties. Another way to put it is that
these objects follow an event handling model, as opposed to a member function
model as in C++. So, methods are activated in response to events. Each
object has a predefined set of events to which it can respond, or the user
may add their own events to this set.

Thus, the object either identifies the event as being bound to a particular
method, and then performs the method and executes the event, or the object
does not know how to process this event and gives back the event unprocessed.
This type of flexible late binding allows for a free-form and intuitive user
environment.


Borland provides a number of products which allow users and developers to
control the behavior of methods.

Object Anatomy and Language Methods

The language used to program a method to an object can vary. One language
that can be used is decision trees, and ObjectVision is the application that
supports decision tree programming. Decision trees are a visual, more
intuitive way to program and can therefore be used by end users, allowing
non-programmers to create applications.

The more sophisticated user, the power user, and a database developer can use
ObjectPAL to add their own methods to these objects. Additionally, with the
large community of dBASE developers creating database applications in the
dBASE language, a method may also have dBASE code bound to an object.

Finally, Borland supports systems developers allowing them to create their
own objects and methods using C++, C, or Pascal. Therefore, Borland is
allowing a range of different users, from the beginning end user to the
systems developer, to create objects that make powerful applications.

Interoperability of Objects

Interoperability occurs at the object level. Besides allowing for
interoperability between applications, Borland's product line for DOS allows
for interoperability at the user interface object level. Users can take a
Paradox for Windows Query By Example (set of objects) and drop them into a
spreadsheet. Another example is to take a dBASE program and make an object
that can be used within ObjectVision. This interoperability at the object
level is a more granular level of interoperability than at the file level.

The mechanism by which objects are moved from one application to another is
through Microsoft's Object Linking and Embedding (OLE) architecture. But, it
is not just Borland applications that can share objects. Because OLE is a
Windows standard, other applications supporting OLE can share these objects,
allowing a mix and match of these objects across different vendors. The fact
that Microsoft has agreed to put object-oriented extensions into Windows is
helpful to Borland because Borland's products have been built with an
object-oriented architecture from the beginning.

The Borland Database Engine

All Borland applications that have an object-oriented user interface will
also include the new Borland Database Engine.

The original Paradox DOS engine architecture is a multi-user, relational
access engine ISAM. Borland also included the Paradox DOS Engine as a part
of Quattro Pro. This allows Quattro Pro users to do spreadsheet analysis
against Paradox data. Furthermore, the Paradox DOS Engine is available
separately as a C, Pascal, or Windows DLL library, so that C++, C, or Pascal
programmers can create programs that access Paradox tables.

In designing the new Borland Database Engine architecture, Borland recognized
certain restrictions on this Paradox DOS Engine architecture:
*Remote SQL database server access can be supported only as an add-on to
Paradox--it is not a part of any of the other products. Accessing SQL data
from Quattro Pro, for example, would require the spreadsheet user to use
Paradox as a gateway to query the SQL database server, creating a Paradox
table first, and then using Quattro Pro through its integrated Paradox DOS
Engine as a second step.
*A corollary to the first restriction is that even though Borland
applications can share common data, they must reside in a Paradox table file
format. The applications can't reside in any other data format.

The new Borland Database Engine architecture eliminates both of these
restrictions.

Object Layer

The Borland Database Engine replaces the Paradox Engine with a virtual layer
that is called the Object Layer. The Object Layer provides a seamless way
for applications to access data. The Object Layer presents a common
interface to each application, independent of that application's specific
file drivers and data requestors. The Object Layer can read and write
Paradox files, dBASE files, Novell Btrieve, and Quattro Pro files directly.
Borland's SQL Link adds mini-computer and mainframe connectivity through the
Object Layer to data stored in Oracle Server and the Microsoft/SYBASE SQL
Server.

The Object Layer is a set of C++ classes that manage database functionality
including cursors, schemas, tables, and the various data formats. The Object
Layer is able to load the appropriate file driver or data requestor,
depending on which type of data the application is attempting to access.

The benefits of the Object Layer mean that Quattro Pro and typical Borland
C++ or Pascal applications can access any of these data formats directly.
Furthermore, it means that all applications can share data across a broad
range of underlying file formats, or data model formats without the
requirement of using Paradox as a gateway between the various types of file
formats.

Query Functionality Added

Borland has taken the Paradox for DOS query engine and moved it into the
Object Layer of the new Borland Database Engine. This should not be confused
with Query By Example, the interface for specifying a query. The query
engine is separate technology that is used to execute query requests.

The query engine in the Object Layer can do very sophisticated queries, such
as joins. This means that an application using the Object Layer can perform
the full range of relational query capabilities, such as set operations,
joins, the Like operator, and scanning for simple matches.

Also, because the query engine is in the Object Layer, the Borland Database
Engine can do query reports. With the Object Layer containing both the
ability to read and write files, as well as the ability to construct new
tables through queries, it can prevent the applications from knowing the
distinction between a query and a file, since both are simply relations.
There are cases where query reports can, therefore, deliver substantial
increases in query performance, and make the whole process easier for the
user.

In Paradox for DOS, the Answer table must first be created by the engine and,
as a separate function, produce the report by re-reading the table. In
Paradox for Windows, reporting on the query output allows the new Borland
Database Engine to take the output of the query and send it directly into the
report generator without creating the temporary Answer table file.
Therefore, having the query engine in the Object Layer can provide
substantial functionality and performance improvements.

Integrity Maintenance Added

Additionally, the integrity maintenance functionality from Paradox for DOS
has also been added to the new Borland Database Engine Object Layer. This is
a significant improvement for the user because it means that functionality,
such as referential integrity, is always maintained. This means that a user
can't delete a record in the master table while secondary records are
dependants. Or, from the other direction, it means a user can't add
secondary records without a master record dependency. The benefit is that
data being fully maintained is much more reliable and safe.

Linked Cursors

The Borland Database Engine also supports the notion of linked cursors.

Linked cursors exist when a user is operating on two or more tables which are
properly related under the normalization rules. Each table has a cursor, but
only one table can be "current" to the user at a time. This means that when
the user moves a position in one table, the cursors in the related tables
also move based on the user's integrity maintenance rules inforce when the
tables were created.

The movement of these cursors automates intuitive functions like one-to-many
forms or one-to-many-to-many forms, so that the user is not required to
support this through programming. Instead, declaring the schema is
sufficient enough to cause cursors to behave intuitively and properly.

Data Model Independence

The Object Layer actually supports two different interfaces. One interface
is a record interface. The record interface provides record oriented data
file functionality to support many functions in dBASE applications. The
dBASE language definition requires the notion of a record, and in order to
fully support the dBASE language functionality a record oriented interface is
available.

The other interface is a set oriented interface. The set oriented interface
is primarily used by many functions in Paradox and by the object oriented
user interface interactive system.

The benefit of these two interfaces is that all application functionality is
fully supported and both of these interfaces can access the same data. It is
not a translation of data, but rather just a different interface to common
data.

Record Oriented Functionality on Set Oriented Data Models

Many of the Borland Database Engine data formats are based on engines that
are set oriented. For example, all of the SQL database servers are set
oriented and don't operate on a record level.

The Borland Database Engine can support a record oriented interface and still
operate on set oriented data models. One of the benefits of this is the
ability to support browsing on SQL database servers.

Borland Database Engine Interoperability

As a part of the user interface, application interoperability is supported
without the engine at the object level. This can be done through a simple
DDE connection or, as mentioned earlier, through rich support of OLE
applications (clients and servers), which allows for a high level of
information interoperability. However, DDE and OLE are not suitable for
large amounts of data sharing. For example, moving a megabyte of
information, would be very, very slow.

The Borland Database Engine meets the need of interoperability with large
amounts of data. Because the Borland Database Engine supports multiple data
formats, users can directly read and write in a variety of different file
formats. In this way data is easily moved from one application to another.

Scalable Database Engine Architecture

A scalable database architecture is defined as the ability for users to move
from a single user environment to much larger systems, with a minimum of
database application modification, a minimum of incremental Research and
Development system programming, and a minimum of installation requirements.
For example, if a user wants to move from a single user file model to a
multi-user file model, you simply move the data files onto a server, and
adjust the directory path names to the data files within the database
application. Therefore, there should be no new programming required when
moving from a single user file model to a multi-user system. This is also
true when moving to more powerful environments to gain speed and capacity,
without any impact on the application.

The Borland Database Engine supports this kind of scalability, and more.
Beyond the multi-user file server system is the client/server model. Today,
much of the client/server support requires custom development work to port
database applications to use SQL database servers. Typically this
development work is done by the customer and not by the vendor.

The Borland Database Engine is also scalable to client/server systems. The
Interbase connection to the Object Layer allows the customer to replace the
Object Layer with the requester. The requester simply takes the requests
that are going into the Object Layer, and re-routes them over the LAN into an
equivalent "Object Layer Emulator" as a part of the Interbase database
server. So, the Object Layer Emulator lets you move to a client/server
architecture without additional development.

The advantage of scalability is that it provides microcomputer to mainframe
scalability, from the stand-alone microcomputer, through the multi-user file
model, all the way to a distributed database environment with databases
running on mainframes, VAXs, Sun systems, and PCs, to name a few. Moving
applications can be done with no new development activity. All of this
functionality can be supported in an engine that stores data which is
accessible to other SQL database servers. This allows Borland applications
to access these servers as well, and not just the Interbase server.



Interbase

Current Interbase Architecture

Interbase is not just a SQL database server. Instead, Interbase is a general
purpose database server. This is because Interbase supports the concept of a
binary language representation interface. Simply put, this is a proprietary
low level language that is used to interface with the Interbase system
kernel. The binary language representation of Interbase is a proprietary
language based on a record oriented model.

The kernel is the part of Interbase that controls database functions like
security, transactions, queries, manages concurrency, logs transactions,
manages the data dictionary, and journaling (which is an alternative to
traditional record locking systems and is a more effective solution for long
duration transactions).

Multiple Language Interfaces

A SQL interface has been built on top of the binary language representation.
The Interbase SQL language dialect is ANSI Level II SQL compliant and it
supports both static and dynamic connections. The SQL interface is only one
of several supported interfaces.

A second interface that Interbase supports is called GDML. GDML is
Interbase's proprietary 3GL. It supports database applications programming
similar to the dBASE language, and has high-level programming methods such as
loops, variables, and user defined functions. GDML also takes advantage of
the kernel extensions, such as a BLOb data type.

On-Line Complex Processing

The benefit of this architecture design is that Interbase can support on-line
complex processing. Long complex transactions can be done effectively. In
contrast, for example, many other servers are tuned for short burst
transaction processing performance. The advantage of having this extra power
allows the user to perform additional functions, such as the ability to run a
report that takes two hours, while also continuing to have other users
accessing, updating (even backing up) the same data simultaneously, as
opposed to locking the tables that the report is using until the report is
completed. This should have tremendous benefits, for example in an order
tracking system. Can you imagine not being able to take any orders for two
hours just to get a report out?

The functionality behind this support is that Interbase uses a journaling
approach that gives the user a snap shot of the data. Other users can
continue to access data, but the report or query will provide an answer of
what the data was at the start of the command. This means long transactions
can be run without blocking access to other users accessing that same data.

Other Interbase Features

Interbase supports a variety of data types including complex data types,
called BLObs, or binary large objects. BLObs greatly enhance the ability of
Interbase to store complex objects built by C++, document management systems,
and other applications that would not be classified as database applications.

Another supported feature is "User Defined Functions", or UDF, written in
GDML that are stored in the database itself. This implementation allows
these functions to become hardware and operating system independent.
Therefore, all operating system platform implementations of Interbase can
execute GDML without source code modification.

One user defined function class is a "filter". Filters are used for BLOb
type coercion. This benefit allows complex data types to be managed
efficiently and reliably. The filter allows for easy access to this type of
data consistently. For example, if an application may need to store a
CAD/CAM diagram, making it available to users on a PC in the AutoCAD product,
but another user on a MacIntosh is using a different CAD/CAM systems.
Accessing the CAD/CAM diagram with a BLOb filter allows both users to access
the same diagram. Since these filters are a part of the Interbase server,
the filters are platform independent and need only to be written once.

Additionally, some functions known as triggers, can start up a routine
running on the client. These are called "event alerters". This is used to
replace a less sophisticated polling system, such as one requiring the client
to look at the database to see if the event has achieved certain values to
start the trigger. With the Interbase trigger, the client can wait for the
trigger to automatically call back to the client when a specified event
occurs.

Lastly, Interbase supports a fully automatic two-phase commit, which allows
distributed data to be managed more efficiently, and for multiple
transactions accessing common data to be handled far more reliably.

Therefore, the Interbase architecture allows for multiple database requests,
and is perfectly suited for server extensions, such as the ability to add
other interfaces.

Future Interbase Architecture Enhancements

Borland is planning an Interbase extension called the Object Layer Emulator.
The Object Layer Emulator is the system that emulates the commands from the
Borland Database Engine Object Layer to Interbase's various interfaces. In
particular, it has the ability to emulate those commands that are common
within microcomputer file model multi-user databases.

An example of this is a logical delete. A logical delete occurs when a user
deletes a record, but the record isn't actually removed from the table.
Instead, the system tracks the record by "marking" it as deleted, but it is
not actually removed. The record stays in the table until the table is
packed. Many applications rely on this functionality, and it is therefore
important to emulate this capability.

Another example is the use of a random goto which reaches a specified record
as designated by a specified or ordinal value. This is also a unique
microcomputer function that is not usually supported by servers.

A final example is the use of expression indices. A database can have an
index based on an expression of multiple fields, a requirement beyond simple
field concatenation.

These new Interbase features can be implemented in a fashion unavailable to
other servers because of the binary language representation architecture.
Other server vendors have attempted to build these transparent connections
from PC-based relational database products to the SQL database server, but
have been limited by SQL. Interbase is not restricted and is able to fully
emulate these features through SQL, especially if the functionality is based
on record oriented access.

Client and Server Enhancements

Borland is planning enhancements to both the client application engine by
designing an Object Layer and, on the server side, by providing an all new
interface on top of the binary language representation that provides access
to the Interface kernel. Because Borland provides products for both the
client and server, we have the option to bypass SQL altogether in order to
have transparent connectivity. Other vendors cannot do this.

As was mentioned earlier, the trigger functionality already present in
Interbase can be used to add extended objects to the Interbase server. Also,
Interbase will have future support for a variety of standard electronic mail
objects to be stored. It is more of a repository for mail, but can also
support a variety of multi-media objects, from rich texts and graphics, to
moving video and sound, that will allow compound documents to be stored, such
as spreadsheets. All of this data will be stored as objects.

The key advantage of this data management technology is for version control,
or journaling. For example, when constructing an advertisement, a company
can circulate the advertisement for approval, collect all the annotations
on-line, and store all this on the Interbase server. It would be easy to
back up over multiple versions of the advertisement to see its various stages
of construction. This also has very important advantages in software
development.



From the Leader in Customer Satisfaction

"We are very pleased about Borland's willingness to listen to us and develop
products and support that meet our needs."
- Bob Marzulli, Vice President, Distributed Services,
Paine Webber, Inc.


Providing leading edge technology and solutions to our customers is our
number one task. However, our commitment to the customer does not end there.
At Borland we are committed to providing unmatched levels of service and
support to complement our product solutions

J.D. Power and Associates just ranked Borland "Best application software in
customer satisfaction in Small to Medium-sized and Large Businesses."*
Company for company, technology for technology, Borland is your best choice.


A Winning Strategy

Borland International database product family provides a new generation of
data and decision support capabilities and flexibility. The full database
product line will provide customers with a full range of software
products--from databases and spreadsheets, to programming languages, and
database server software--designed for the application development needs of
the 1990s. And, each application will be designed to work together with the
others, seamlessly in the graphical environments.

Borland's database strategy adds up to a "win" for database customers. The
best products, the best technology, and the best support, from the company
that guarantees customer software investments now, and in the future.

__________________________________
* 1991 J.D. Power and Associates Computer End User Satisfaction Study: Phase
I. Office-Based Small to Medium-sized Businesses.SM Response from Business
End Users at 1,784 business sites. Small to medium-sized businesses were
based on office sites with between 1 and 499 employees. 1991 J.D. Power and
Associates Computer End User Satisfaction Study: Phase III: Office-Based
Large Businesses.SM Responses from Business End Users at 1,094 business
sites. Large businesses were based on office sites with 500 or more
employees. J.D. Power and Associates is a service mark of J.D. Power and
Associates. Borland is a registered trademark of Borland International, Inc.
Copyright 1991 Borland International, Inc. All rights reserved.