Horizontal/Vertical Develpoment Verification
I managing two projects as Project Manager.
One, being developed in more a vertical way. Sure, there are layers, but each new aspects added by subsequent development task requires simply enhancing each of existing layer (component) with a part supporting added functionality.
Another project is developed in a horizontal way. It has been designed and planned in such a way layers can be and are developed independently .
First approach is ideal, when I want to perform end-user tests (e.g. shows working software to executive) or even perform User Acceptance Tests during software development.
Second approach does not support this kind of tests unless the layers get integrated in the end of the development. This is easy to make unit tests (tests of separate layers, components), but still the level of technical details this testing requires makes verification of output hard for me, i.e., for Project Manager.
1. How do you verify whether the output of development of your programmers does what it expected to do without going much into low-level details?
2. How would you verify that in the second approach?
Re: Horizontal/Vertical Develpoment Verification
I'm not sure I'm reading this correctly so forgive me if I'm going off in an odd direction.
In my organization we have two very large projects that are managed by different groups, developed separately and released on their own schedules. However, there are a few places in enterprise applications where they "touch", which presents a challenge for testing. We've tried a few things, but our ultimate goal is to isolate the released (gold-standard) version of each to a shared test server, which can be upgraded with one project's release candidate so it can be tested against the live version of the other.
Might it be possible for you to follow the same process with the independent layers of the horizontally-developed project?
Also, everything has an end user. If you are struggling with the technical details, you might be employing the wrong users to help you test. If the end user is another system, then that's the system that can be tested for the public (if it's sexy enough for anyone to care to see it).
I hope this helps a little, at least!