> I recall hearing that it is not necessary to recompile a program
> just because a new item has been added to a dataset.  That is of
> course that the program declares each item being accessed and
> that "all-list" is not used on the affected set.   The dispute is
> that you must recompile regardless of how the list is defined in
> the program.

You are correct.  Recompiling a program will have *no* effect on the
resulting program or the database, unless some part of the source code for
that program has changed.  Now if you have a COBOL COPYLIB or some $INCLUDE
file that describes the layout of the dataset that you have updated, then
recompiling would cause the program to use this new definition.  Whether you
need to do this depends on the change to the dataset.  As you say, if you
are using only explicit item lists (and using item names rather than item
numbers) and possibly the "*;" list, then your old programs should be
compatible with the new database.

Of course if your old program PUTs to the dataset, then it will not be
including the new item and that new item will get its default value
(basically binary zeros) which for an X type field results in null
characters in the field rather than spaces which might be a problem.

If the field you have added is a key item (search or sort), then there could
be other impacts as well.

All of the issues should be fairly obvious if you think about it for a
while.  There are no hidden things going on behind your back (as there are
with many SQL databases, etc.) that you need to worry about.

HTH,

Gavin