Bad practice? Technically? I think not - the underlying structure tables 
and files is the same - just do a DMPOBJ of an example of both, and it 
is obvious. The underlying MI object types are the same - *FILE, *FMT, 
*MEM (member), *QDDS.
Here are a couple examples of DMPOBJ output of a file and a table that 
are otherwise the same -
OBJECT TYPE- SPACE                                           *FILE
NAME-        TESTPFDDS                       TYPE-          19 
SUBTYPE-          01
OBJECT TYPE- SPACE                                           *FMT
NAME-        TESTPFDDSR                      TYPE-          19 
SUBTYPE-          51
OBJECT TYPE- CURSOR                                          *MEM
NAME-        TESTPFDDS TESTPFDDS             TYPE-          0D 
SUBTYPE-          50
OBJECT TYPE-           DATA SPACE                                      
*QDDS
NAME-        TESTPFDDS TESTPFDDS             TYPE-          0B 
SUBTYPE-          90
OBJECT TYPE- SPACE                                           *FILE
NAME-        TESTPFDDL                       TYPE-          19 
SUBTYPE-          01
OBJECT TYPE- SPACE                                           *FMT
NAME-        TESTPFDDL                       TYPE-          19 
SUBTYPE-          51
OBJECT TYPE- CURSOR                                          *MEM
NAME-        TESTPFDDL TESTPFDDL             TYPE-          0D 
SUBTYPE-          50
OBJECT TYPE-           DATA SPACE                                      
*QDDS
NAME-        TESTPFDDL TESTPFDDL             TYPE-          0B 
SUBTYPE-          90
There will be some values in the first few rows of the dump, which hold 
the object-based attributes that distinguish them.
Or just do a DSPFD - the main section at the beginning will have almost 
everything the same - one difference is SQL attributes in the DDL-based 
objects.
So I feel that advice that DDL-based objects with RLA is bad - that is 
more of a religious argument - yes, there are are possible performance 
advantages due to larger block sizes in indexes, but that, for me, means 
that using an index with native IO is even better than using an LF. It's 
more of taking a position to move to SQL only - and that doesn't have 
technical merit, so much as it might have merit as to aligning with what 
is done if you work with another RDBMS.
Regards
Vern
On 1/14/2017 11:43 AM, D*B wrote:
 SQL and RLA compliment one another rather than compete.
creating tables with SQL DDL and using RLA, as some experts recommend, 
is a bad practice, you are falling bacl behind DDS defined join files. 
tumbling around in the database to grab together data, a join 
operation would deliver easily. Another weakness of RLA is: the tight 
coupling from database implementation and application.
-snip-
As an Amazon Associate we earn from qualifying purchases.