OBJECTIVE OBJECTIVITY? REPLY TO GUZENDA

The editor of my (now discontinued) Against the Grain column, has recently notified me that "in the interest of fair play" he published a so-called "article" The ODBMS lives! A rebuttal to Fabian Pascal, by Leon Guzenda, the Chief Technology Officer of Objectivity. He added: "I would discourage you from writing a column in response, since those sort of "tit-for-tat" pieces aren't very interesting or popular."

Funny that: the media bends over backwards to be fair to vendors and publish their "rebuttals", even if they are not rebuttals at all, but sales pitches, but responses that would expose those for what they really are "not very interesting or popular”. And then they have the nerve to call it “balance” (see On the Trade Media’s "Balance").

There’s nothing of substance to rebut in Guzenda's short sales pitch that a reader has not already expressed in his reply (the name of Guzenda's company is rather ironic in this context.) Nevertheless, I am taking the trouble to respond to it here to demonstrate yet again:

· lack of knowledge of fundamentals by no less than Chief Technology Officers of DBMS companies (see also What Do You Mean, "Post-Relational"?, Comments on an Interview with Jim Gray , and Unstructured Thinking.) What, under these circumstances, can then be expected of the average practitioner?

· the inability by even technical vendor personnel to address substantive logical issues without digressing into irrelevant implementation details and unsubstantiated marketing claims

Guzenda states that "Much as I respect Fabian Pascal's views on databases, I differ with him on the merits of ODBMSs. Everything he says is a regurgitation of the RDBMS vendors' early "Fear, Uncertainty and Doubt" messages to their loyal disciples." [emphasis mine]

Well, first of all, if he respects me for the views that he assigns me, then he certainly should not, because they are not mine, and they are not respectable at all. I dare anybody to produce any material of mine that can be interpreted in the way Guzenda does. Rebuttals require the rebutter to know and understand what he purports to rebut. Not only does Guzenda fail on both counts, but he does not rebut even his own straw man that he sets up, one that has nothing to do with my positions. And get this: he accuses me, of all people, of "regurgitating RDBMS vendors"!! What (object) planet has he been living on? Coming from a vendor pushing a technology he is vested in, without any substantiation, this is real rich.

Given his introduction, one expects Guzenda would producethe specific merits of ODBMSs relative to RDBMSs. But does he?

"Much of [Fabian's criticism of ODBMS] is no longer true (Objectivity/DB doesn't use pointers and never has, for instance)"

Object IDs are pointers! Albeit logical, not physical pointers, but pointers nevertheless. If Guzenda does not realize this, then he does not understand the technology that he purports to be an expert on.

"[Fabian] ignores a few simple facts. ODBMSs targeted tasks that relational databases were and still are bad at handling."

But there are no relational DBMSs (not databases!). What Guzenda is referring to are SQL DBMSs which, as I have been stressing over and over again, are not the same thing; a technical officer should know that.

Whatever "tasks" existing products are "bad at handling"—and Guzenda does not specify any—this has nothing to do with their being relational (which they really are not)! The real question is what does "handling" mean—it's a conveniently fuzzy term which, when probed into, the argument behind it evaporates (see Chapter 1 inPRACTICAL ISSUES IN DATABASE MANAGEMENT on one of the “tasks”,

so-called “complex data types”).

"We've also overcome the "They won't scale" argument. The biggest database that uses Objectivity/DB currently stands at 578 Terabytes, growing by roughly a Terabyte per day. We believe that it is bigger than Oracle's top 200 sites combined! We also have telecom customers that concurrently service hundreds of thousands of customers per set of fault tolerant equipment."

Say what? Where did I ever say anything about scaling? Scaling is a physical implementation aspect which has absolutely nothing to do with the data model, which is a purely logical construct. It is completely irrelevant in any discussion about the merits of data models underlying ODBMS (if any!) and RDBMS. Both ODBMSs and RDBMSs can be implemented efficiently or not. What such arguments reveal is that ODBMS proponents are confused about the logical and physical levels and cannot keep them distinct in their minds (see The Logical Physical Confusion.)

As Chris Date has pointed out so many times, the so-called "object-model" is much closer to physical storage than its proponents realize, which weakens further—if that is possible—its status as a competitor of the relational model. (Incidentally, can Guzenda specify— precisely, please—the “object model”’s structure, integrity and manipulation equivalents, and its theoretical underpinnings? What does it substitute for predicate logic and set theory?)

"Have you ever tried online schema migration with an RDBMS? It's easy with our Active Schema package. In short, a well constructed ODBMS can be faster, more reliable, more scalable, easier to use and cost less to maintain than its relational equivalent."

And he accuses me of regurgitating vendors??? Is this Guzenda's explanation of ODBMS advantages? He simply declares superiority, without providing any evidence. Now, that's quite easy. I could simply replace ODBMS with RDBMS and say the same thing: a well constructed RDBMS can be faster, more reliable, more scalable, easier to use and cost less to maintain than its object equivalent. On what grounds should the reader decide who is right? How does Guzenda know his argument is true, given that there are no relational DBMSs for anybody to have tried schema migration on? Had he really known and understood the relational model, he would have known that one of its main objectives was precisely to maximize flexibility in the face of change.

“I should also mention that we also support ODBC and SQL so that Fabian and his peers can use their favorite tools. If they really want to regard a network of objects as a clumsy bunch of tables and JOIN table constructs, they can. It will just run slower than if they had done it the elegant ODBMS way."

As the reader rebutting him has pointed out, the notion that SQL is my "favorite tool" is utter nonsense. As to the rest of his argument, I don't know whether to laugh or cry. Shouldn’t Chief Technical Officers, responsible for the design and implementations of DBMS products, be expected not to confuse the logical and physical levels? Is there any wonder that the industry and its products are in a such sorry state?

"The RDBMS vendors have moved from an "Objects are irrelevant" to a "We have objects too" pitch."

So what? SQL vendors are vendors just as Objectivity is. In the absence of knowledge by both vendor personnel and users, they all ride various fads, with disregard as to sound foundations. The fact that Oracle or IBM use object terminology in their marketing, or bastardize their products with object features to compete with the likes of Objectivity is not proof of the merits of ODBMS.

"XML is becoming increasingly important in the software industry. The most efficient XML databases are ODBMSs with support for an additional language. The RDBMS vendors have added XML, but their offerings are noticeably inferior to the XML-on-ODBMS offerings from companies like X-hive, an Objectivity partner."

Guzenda grounds his arguments not on functional or technical merit, but on what the industry does; marketing notwithstanding, at the very least he should be able to distinguish between them. The problem is precisely that the industry does most things wrong precisely because of the kind of thinking and knowledge Guzenda displays.

There is hardly a better example of that than XML. Invented by text publishers without any understanding of database management, it is essentially a relabeled rehash of the old hierarchic DBMSs, which we got rid of decades ago (see my several Against the Grain columns on XML).

Even if one ignores that, can Guzenda say--precisely, please--on what grounds does he claim that RDBMSs (not SQL DBMSs!) are inferior to ODBMSs with regard to support of XML as a domain? (again, see Chapter 1 in my book and Unstructured Thinking). In fact, does he even know what a domain is and how it relates to a data model (see On Suneido “DBMS” and Data Types.)

"Sure, you can keep adding layers to your old relational engine, but if I offered you a 50 miles per gallon F-16 for the same price as a Chevy truck with strapped on wings and a bigger engine which would you rather fly? It seems that Fabian would prefer to stick with the Chevy. So do a lot of IT managers. It takes time to change attitudes and to sell a new technology into a company that has already written off the cost of its RDBMS site license, but the ODBMSs and XML databases are steadily gaining ground."

Is this a technical argument on merit? What “old relational engines”?

The editor told me that Guzenda’s was the “only coherently written rebuttal he received” to my OO columns. If so, there is no better evidence for the sad state of the database industry.

Now, you can argue that “the market” has decided that Guzenda knows enough to hold a technical position, and that’s what counts. But that’s only because the market does not know any better, not because it does.

Posted 11/6/02