| 1 |
Design Considerations for TriBITS |
| 2 |
--------------------------------- |
| 3 |
|
| 4 |
Some of the basic requirements and design goals for TriBITS are outlined in |
| 5 |
the `TriBITS Overview`_ document. |
| 6 |
|
| 7 |
.. ToDo: Discuss design requirements for TriBITS in more detail. |
| 8 |
|
| 9 |
As stated in `TriBITS Dependency Handling Behaviors`_, `No circular |
| 10 |
dependencies of any kind are allowed`_. That is, no TriBITS package (or |
| 11 |
its tests) can declare a dependency on a `downstream`_ package, period! To |
| 12 |
some, this might seem over constraining but adding support for circular |
| 13 |
dependencies to the TriBITS system would add significant complexity and |
| 14 |
space/time overhead and is a bad idea from a basic software engineering |
| 15 |
perspective (see the *ADP (Acyclic Dependencies Principle)* in `Software |
| 16 |
Engineering Packaging Principles`_). From a versioning, building, and |
| 17 |
change-prorogation perspective, any packages involved in a circular dependency |
| 18 |
would need to be treated as a single software engineering package anyway so |
| 19 |
TriBITS forces development teams to glob all of this stuff together into a |
| 20 |
single `TriBITS Package`_ when cycles in software exist. There are |
| 21 |
numerous wonderful ways to break circular dependencies between packages that |
| 22 |
are proven and well established in the SE community (for example, see [`Agile |
| 23 |
Software Development, 2003`_]). |
| 24 |
|
| 25 |
.. ToDo: Discuss why it is a good idea to explicitly list packages instead of |
| 26 |
.. just searching for them. Hint: Think performance and circular |
| 27 |
.. dependencies! |