George:

> I've sort of been following this thread and the question of whether or not
> an auto-increment column is a major improvement or not.
>
> First, I think whether or not it's major is simply a matter of
> perspective. What one person sees as being a big deal, may seem trivial to
> someone else.

Agreed.


> Second, and I think more importantly, no one has touched on what this
> allows the developer to do that he/she couldn't before.
>
> The main reaction of folks who already have routines to handle the
> generation of PKs seems to be, "Well, I already have something like that
> and I see no reason to change to using an auto-increment column."
>
> I think that the folks that have expressed might want to have some second
> thoughts. Let's assume that this is the case, and, after the release of
> Toledo, developers are porting 7.0 applications to it. Let's, however,
> consider another possibility: That the application may in the near future
> need to be up-sized to SQL Server 7.0 or 2000.
>
> Both of those support identity columns. Identity columns are essentially
> the same thing as an auto-increment. Further, I think it's a reasonable
> guess to make that the Upsizing Wizard will be modified so that auto-
> increment columns map to SQL Server indentity columns.
>
> Wouldn't make a good deal of sense to dump the Primary key generation
> routines in favor of an auto-increment? If one did, not only would there
> be less code to currently maintain, but also the upsizing process would
> more seamless and less modifications required, especially if the developer
> had the good sense design against the Fox tables as if they were SQL
> Server tables.

> That's my take.

Very true if you know that any port will be done to a RDBMS that supports
auto-increment columns.

On the other hand, not all RDBMS support auto-increment columns. You
will have to write something somewhere to port your application if it is
currently using this VFP feature. You may not have to do anything other
than setting the same table in the new RDBMS if you already have a function
that handles the primary key values generation process.

I never know which RDBMS I will be using next and I chose the more generic
approach based on this reality. I am not saying that I would choose the
auto-increment columns path if I knew all my ports would be to RDBMS that
support the feature; all I am saying is that auto-increment columns are
not a good option for my reality (and to give peace a chance).

Your reality and conclusions may be different than mine. Nothing wrong
with that.

Daniel