RE: ODBC (was RE: Green screen - it's time is over )

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> -----Original Message-----
> From: Steve Richter
>
> Does a logical file on the server, with explicitly declared flds, used as
> the target file name of the client's odbc calls,  solve the column name
> dependency concerns?
>
> Dont know much about ODBC, but if the odbc connection info ( actual target
> system, library and file name ) can be stored external to the client pgm,
> would that satisfy the "client code is dependent on the target
> file name and
> location" objection?

Perhaps, and perhaps.  I am not saying that all ODBC is bad; I'm saying that
it's more susceptible to changes in the database layout than a server-based
environment.  That doesn't mean you can't avoid some of the problems with
stringent coding guidelines and clever techniques.

The fact is that you can write perfectly good code using just about any
language, provided you have some strict guidelines in place.  Believe it or
not, you can write flexible, maintainable code in assembly language.  While
you've pointed out a couple of ways to insulate the ODBC programmer from
changing database characteristics, what percentage of ODBC access is written
that way, do you think?  My guess is that the overwhelming majority is not.

But each guideline adds its own complexity as well.  How do you store the
"external connection definitions"?  How do you access them?  How do you make
sure that all your clients are getting the same definitions?  Are they
stored on the host?  Which host?  As you start adding layers of complexity,
the "ease of use" of ODBC evaporates and make the client/server approach
even that much more appealing.

And even with all the additional work required, this "enhanced ODBC"
approach only solves some of the problems.  Take a look at my earlier post,
where I outline a simple, real-world situation where shipment information
was moved from one level of the hierarchy to a new level.  A server-based
architecture takes this in stride much better than an ODBC architecture.

Joe Pluta
www.plutabrothers.com