RE: MVC Architecture and transactions

From: Buck

Joe Pluta wrote:

Buck, you're missing my point again! That order that you rolled back

should

be posted with a timestamp that you fetched at the beginning the update,

and

then the order would have a timestamp prior to the update of the master
file.

That's the header, not the detail.

Okay, maybe I'm missing something. If you're updating an order, I can't see
that the update of one detail record is more than a few milliseconds after
the update of another, but that's me.

It just seems to me that you've unnecessarily convoluted this.

CC is easy to implement, easy to understand and just plain works.
That's my opinion, and the reader can take from that what they will. I
don't see a need to caution people away from CC, rather the opposite. I
felt that adding an actual developer's experience in a demanding
performance regime would be useful.

I agree more or less with all the above, but it is NOT free, and I don't
like recommending any technology unless it's necessary. That's just the way
I am, I guess <smile>.

I've been forced to look at my experience and reflect on whether it was
justified or my desire to tinker with something new. After reflecting
on it, I can say with honesty that I would solve the same problem with
commitment control today. That doesn't diminish your addition to the
thread one whit in my eyes. Thanks for your time and your thoughts.

And I'm certainly not implying that your decision was wrong. In fact, two
people can reasonably see the same situation and design different solutions.
I just wanted to make sure all sides of the discussion were addressed.

Joe