WHEN IS BEING AN OBJECT A GOOD THING?
Databases, Objects and the Future

by Denise Caruso

c. 1990 Media Letter

Writing about databases is an act of raw courage, since the topic is instant 
Sominex to most of the population. But as we move forward into a world 
where heterogeneous network servers brim with video, still images, sound, 
animation and graphics, a database or its functional equivalent will be central 
to a company's ability to exploit the power of its vast stores of electronic data.

The entrenched technologies, flat-file and relational, are at least adequate in 
handling multiple data types. But the concept of object-oriented 
programming in general is becoming popular because defining types of data 
as "objects" makes prototyping and writing software faster and more flexible. 
As great strides are made toward the widespread use of objects, new 
technologies make it easier to use them in databases and other applications.

In the classical sense, an object is equal to the data itself (i.e., the picture, 
sound, paragraph of text, spreadsheet cell, whatever) plus the methods which 
define how it interacts with other objects and vice versa. It takes far less time 
to write a program that manipulates such an object than it does to get into the 
trenches and, for example, rewrite hirsute user interface code every time you 
want to create a new application. That's why NextStep's Interface Builder 
wins such high praise in the software development community. A reusable 
library of objects shaves an enormous amount of time off a product's 
development cycle, time which can always be better spent cooking up new 
applications or features.

AND WE LIKE IT, TOO 

In multimedia, object orientation is especially valuable because by definition, 
most objects used in multimedia applications are complex and compound in 
their structure. In addition, there are tons of them, and they have to work 
closely together -- very often, video will be accompanied by sound, maybe text, 
maybe a spreadsheet cell, maybe a combination of all three, or new ones that 
haven't even been derived yet. The comparatively simple rules which 
govern relational and flat-file databases are not likely to be robust enough to 
handle the kind of sophisticated data maneuvering required in future 
multimedia applications.

As a result, a growing number of technologists concerned with such things 
believe that what are thought of today as database functions -- storage and 
retrieval, as well as sophisticated data manipulation -- will actually become 
part of the underlying structure of either an application or a computer's 
operating system, or both.

One means to this end is to define standard data types as classical "objects" 
(that is, data plus methods, as described earlier), then provide a layer through 
which they communicate with an application, a kind of brain trust of object 
intelligence that knows what and where they are and what they are able to do. 
This layer, sometimes called a "library," will ideally sit between the 
application and the file system, and provide a greatly simplified means for 
developers to access and manipulate, within their applications, any stored 
object.

Though it's stretching the metaphor until it screams, picture the benefit of 
such a layer as the difference between ringing the butler to fetch a glass of port 
vs. going to the bar, opening the cupboard, rummaging around for the port 
bottle, finding the corkscrew, uncorking it, picking a clean glass, pouring the 
wine, and heading back to the parlor to finish watching "Cop Rock." It's 
simply the difference between "ask and you shall receive" and doing it all 
yourself.

HOW YOU LOOK AT IT

Though a wave of new object-oriented databases is breaking today, only one 
such company is using the "layer" approach to application development: 
Versant Object Technology of Menlo Park, CA. Versant, which has been 
shipping its object database since June on Sun Microsystems workstations and 
IBM's RS/6000, is readying just such a layer of software called Versant Star, 
which sits between the database and its applications.

Star, says Versant CEO Mike Seashols, is a "universal communications layer" 
that takes the protocols generated by applications and interprets them for 
whatever data retrieval system is being used, whether object-oriented or flat-
file or relational. "Our position is to take an enabling technology and apply it 
to the problem of integrating information -- records, text, objects, images and 
spatial data -- regardless of source," says Seashols. "Not even IBM can define 
how you lay data down on a disk, a central data dictionary and the 
architecture of getting data in and out of it. This is the first real technology on 
the market to start making effective integration happen."

Patriot Partners, the new venture formed by IBM Corp. and Metaphor 
Computer Systems of Mountain View, CA, has an even larger picture in 
mind. Using the same object-oriented approach that Metaphor has 
successfully used in its Data Interpretation System (DIS), Patriot will spend 
the next two years crafting an application development environment for 
OS/2 and UNIX-based machines based on a layer of technology similar in 
function to Versant's, which will grant all software developers -- not just 
those writing database applications -- easy access to a large library of objects.

Both Patriot and Versant, then, are approaching the problem in a similar 
fashion -- make the use of objects, and subsequently the porting of 
applications, an elegant no-brainer for developers. However, a slight 
difference in approach aside, both companies are being fairly proprietary 
about who they serve -- Versant by working only with database applications, 
and Patriot by its choice of platforms.

WHO NEEDS STANDARDS?

But one new company called Common Knowledge, based in Saratoga, CA, 
and founded by Ann Arbor Softworks founder and former Ashton-Tate chief 
scientist Scott Wiener, is taking a broader approach. Wiener, whose team cut 
its teeth handling databases and multiple data types for its work on both 
FullWrite Professional and at A-T, is working on a product designed to be a 
replacement for any application's file system which will reside on a network 
server and support any computer, operating system or application.

Called the WaveServer Access Library, Wiener says it's a kind of database for 
applications -- not users -- to talk to. Sitting between the operating system and 
the application, WaveServer is both a server pool sorting the stuff 
applications are interested in -- paragraphs, spreadsheet cells, graphics, 
digitized sound -- and the communication between the pool and applications. 
Thus applications actually talk to a smart server as opposed to "dumb, passive 
files" as Wiener calls them.

What makes the server smart is the way WaveServer describes the objects 
within the pool, a method for which the firm is seeking patent protection. 
Each object has a dictionary for indexing which includes both its name and its 
value -- which can include anything from a keyword to another name to a 
group to "has anything changed?" rules, and ad infinitum. Applications 
query the pool like a user would query a database, and receive data based on 
anything included in an object's dictionary. The code overhead for such 
indexing is small enough that Wiener believes it will work on a 1MB 
Macintosh.

Designed to work with any network (AppleTalk, TokenRing, whatever) with 
nodes (Mac, PC, UNIX boxes, it doesn't matter) and a server (Mac, OS/2, 
UNIX, Vax, IBM 390, you pick), WaveServer is especially designed to ease the 
pain of groupware authors stymied by the wide range of operating systems 
and hardware found on large networks. "Network complexity is just not the 
kind of stuff that applications developers want to deal with," he says. "We 
need to get rid of the file system and get down to objects. If it were still 1960 
and the world was all ASCII, it would be okay. But it's not."

WaveServer, at least in concept, has been greeted with open arms by many of 
the hardware companies which Wiener has approached about the project. He 
hopes it will eventually be built into the operating systems of such firms as 
Silicon Graphics, Apple, Sun, NeXT, IBM and Go.

In addition, Wiener is already dealing with hand-picked software developers 
to help his team crystallize the product's final features. "We picture people 
integrating in cycles -- it takes six months, so we want to start talking to them 
now," he says. "The ideal case would be that we become a technical 
development company and license this technology to major platform players, 
so every applications developer could count on it being there. It's the Adobe 
position -- to be the 'standard' for data storage in the same way Adobe is the 
standard for page description."