RE: Logical File or OPNQRYF or any other way ? - Legacy

Charles & Rob,

No, Joe's reply does not "cover"(refute) what I said; not by a long
shot.   It just simply shows Joe's blind prejudice toward the dogma of
Rochester and their iSeries/AS/400 and his penchant for ignoring or
otherwise not understand the reality of the true relational world.   He
is right about one thing though, there is probably no "true" relational
database engine according to Codd's rules or their interpretation by
Chris Date or others that were close to Ted and understood the essence
of his rules.  My use of the word "true" was a quick label to
differentiate between what is consider by the real world as "true"
relational (the other DB2s, lead by DB2/MVS and others like Oracle).   

Again, the bottom line is not only what is considered relational from
the perspective of the essence of Codd's rules but, somewhat more
importantly, what the commercial world thinks is a true relational
database and how the Rochester implementation in the S38/AS/400/iSeries
does not compare nor compete when IT buyers are considering a purchase
of a relational platform.   Again, I think the iSeries, etc., can
compete if Rochester will move more rapidly toward the architecture of
DB2/MVS.
It would not only make the iSeries truly in lock-step with the DB2
family, it would allow it to be taken seriously by those considering a
relational platform.  

You said "Let me add my observations: the problem isn't DB2 for
iSeries, as DB2 for iSeries is as relational as any other RDBMS out
there.  The problem is most databases found in DB2 for iSeries aren't
relational.

For a relational DB to be managed by a relational DBMS, there have to
be relationships between the data defined.  All too often, DBs on the
iSeries don't have primary and foreign keys defined."

I agree with the part of your first statement "The problem is most
databases found in DB2 for iSeries aren't relational. "   But, of
course, I don't agree with the first part.  One of the problems is
iSeries folks still thinking that creating their legacy flat-file
structures makes them somehow relational because Rochester leads them to
believe they are because of the claim the OS/400 is truly relational. 

I don't quite agree with your second statement.  As others have pointed
out you can have TABLES defined in a relational database with no
relationships nor with any keys and yet they are still considered
relational tables if they are managed by a true relational database
engine that adheres to the essence of Codd's rules.   I may argue the
interpretation of Codd's rules in a later note.   Some of the debate is
academic but the essence of what Ted meant is what makes database
engines like DB2/MVS/VM, etc, and Oracle considered truly relational and
the iSeries not.   What I've said earlier about what's considered the
true relational world of DB2/MVS, Oracle and the like and the
non-acceptance of the iSeries when considered by that world, is more
important.   

So let's bring things somewhat full circle:   My definition of the
legacy iSeries file structures being flat (1st normal form) when
compared to such structures as true relational, or IMS or the like, is
correct.   I'll certainly accept someone's change of my definition that
most files for most applications built on the iSeries files are
non-relational.     

Rob,  

"That sounds like a lot of theory.  But can you name one actual
difference that a developer might notice?"   

How about only being able to use SQL to get anything from the
relational database.  There is no back door using any other access
control verbs or method.  If you are a legacy application developer this
will be considered a bad thing.  If you are used to the modern world of
relational, this is a given of how you access true relational databases.
   

It may be that Rochester and devout followers don't really want to be a
part of the rest of the "norm" of application development world.  If the
"uniqueness" of the iSeries was an attribute that was in demand by the
majority of the development world, then I'd say its future is assured. 
But, that doesn't appear to be the case if you look at the whole world
from PCs to mainframes.   I think it CAN be a viable platform for the
future but its followers and developers are going to have to change in a
couple of areas; its claim to be a relational platform when compared to
the standards of rest of the DB2 world, Oracle and the like and its
steadfast insistence in using RPG and not a language that is a part of
the mainstream.   I THINK Rochester is starting to see the light
although slowly probably so it doesn't cause a riot of the devoutly
religious followers.  I'm hoping the followers will help lead the way
which is what I'm trying to do.

Sincerely,

Dave Odom