Re: Sorting records in Physical and Logical files

Walden H. Leverich wrote:

I disagree. What we had (and still have if you use RLA) was a procedural
programming model. We had to tell data-management "how" to get the data
we wanted. In the new (SQL) model we tell data-management "what" we want
and leave it up to data-management to figure out "how" to do its job.
This is no different from the previous advances in storage management
(remember clusters heads and sectors) or communications (RTS/CTS
management).

Oh pishposh. SQL is hardly "new". It's nearly 40 years old, which is really quite old in programming terms. It's certainly a *different* model, but even the concept of declarative vs. imperative is hardly new. SQL is better for some things, procedural RLA better for others.

The biggest issue with SQL (and indeed anything where you abdicate control of what the computer is doing) is that you have to rely on the computer to make good decisions. SQL is essentially a powerful expert system for database access, and its decisions are only as good as the combination of two things: the programmers who programmed the expert system and the information it has to make its decisions.

So while you're right that SQL is more of a "what" than a "how" (declarative vs. imperative) that doesn't make it "better" than procedural, and while it may be "newer" it's only marginally so. There's a reason that ISAM (a/k/a RLA) has survived alongside SQL for these past three decades: it's a really good tool <smile>.

In fact, ISAM is what made the PC into the business tool it is today: you may remember a little package called dBase.

Joe