Re: Database I/O Modernization

On Fri, Sep 8, 2017 at 5:57 PM, Roger Harman <roger.harman@xxxxxxxxxxx>
wrote:

Unless I'm missing something, I don't see where it shields you form record
format changes. Looks like it's just using a pointer to a DS for the data.

You're right on both counts. The service program would need to be
recompiled in order to incorporate record format changes and avoid record
level checks. The pointer that is passed in the parameter references an
externally described data in the caller. Therefore any program that calls
the I/O service program would need to be recompiled after record format
changes.

To quote Creedence Clearwater.. I see trouble on the way.

I see trouble too, if you recompile the service program to avoid a level
check, but fail to recompile the programs that reference the external data
structure, Callers would likely insert invalid data into newly added
fields.

Funny how we thought (original) free form was pretty modern - until you see

it like this nowadays.... Thank you Barbara Morris & friends for **free -
I'll never look back.

I question the relevance of that. One could change the source code to be
totally **free, but it seems to me that discussing the pros and cons of the
interface would be a lot more pertinent.