[AccessD] Access versioning / tracking changes

John, there are some concepts that you need to understand; version and
revision.  Revisions are incremental changes since the last offical
release or version.  Rollback "destroys" previous revisions.  You can
always checkout an earlier revision without destroying the other revision
but, you still wind up with lost changes.

SCM tools operate on text-based files or source code.  They have a
built-in ability to "compare" any two revisions for changes and display
the results, similair to standalone that compare text files.  Access is
binary so you can not compare anything.  That is your problem.  As others
have mentioned there are some plug-ins available that could extract the
tables/queries/reports but it still binary.

> I use a somewhat simple two table change request database for tracking
> changes to my Access
> projects.  I have to admit I find it problematic to track changes to a
> level that allows backing out
> any specific change while leaving the rest.  This has resulted in "rolling
> back" to a specific level
> when a problem comes up.  And yes they do test but things do happen.
>
> I have a CR table where the client places their change requests with
> explanatory text.  It has the
> typical requested date / requesting person / date to test / date tested
> etc.  A child table holds
> what I do with explanatory text and a test regimen to test that it work, a
> text for what they found
> in test (if problems).  I can add another record as a response to that
> testing problem etc.
>
> The problem I run into is that any significant change may involve a change
> to N queries, additional
> fields or entire tables, code modules and so forth.  A change may be
> trivial or it may be an entire
> subsystem.  I have never found a way to really document in sufficient
> detail what I did to implement
> the change that would allow me to back out just that change, at least of
> the change is very complex.

Never gonna happen.  You will still loose everything after that change
provided that there are later revisions.  Even with text-based source code
you can not just get a portion of a revision, it is the whole revision or
nothing.  A possible solution would be to commit one requirement at a time
instead of several.  Still will not solve your problem.

>
> If I get two or three changes in and then one four changes back is found
> to be a problem such that
> they want to roll it back, I often times cant.  If we roll back all the
> changes since (go back to a
> previous version) then we lose all of the actual work done since.

Can not be helped.  Then either the testing was incomplete or the
requirements were faulty or the understanding of the requirement was
faulty.

>
> I have never worked in a large design team and witnessed how this is
> generally done.  I am wondering
> how you guys handle this stuff.  Any words of wisdom?  Tools?  tips?
>
> Can we have a discussion on this?

You should have some requirements that the customer must provide testing
parameters/data so you can test and understand their changes.  If they
decide that those changes do not work then they must understand the risks.

Mike...