The Reference Architecture defines that preferred platform stacks for our future strategic efforts. Our architectural and build plans will use the RA to influence strategic approaches to platform considerations, particularly when it comes to meeting non-functional requirements.
The "tiered platform" support is more tactical in nature. The basic intention here is to set clear expectations between product management, channels and AD about the testing approach for the various options that exist in the platform matrices. In areas where there are comparable choices, the most preferred or recommended option will get the Tier 1 support. Other choices that are not as preferrable become tested in lower tiers of support.
The basic tiering philosophy says that:
Tier 1 gets fully tested for functionality on every release made GA.
Tier 2 only gets regression tested as part of making releases available.
Tier 3 only gets tested when specifically requested.
As an example, BASE24-eps version 11.1 will support two different databases when running on System P (AIX). One of those databases is DB2 and the other is CTree. Which one is preferrable? It depends on who you ask. If everyone agrees that these are exactly equal in preference and should each be tested fully with any effort we undertake, than both of those options will recieve Tier 1 support. We have been making some motions to reduce our support of Ctree over the long term. If that continues to be the case, we may get Product Management to define the combination of P-series/DB2 will be our supported as part of our Tier 1 testing, while P-series/Ctree would have a drop to Tier 2 or Tier 3.
I consider this initiative to be setting the ground rules around what the tiers are. Once that is established, product management can then define which level of support they want to assign to a specific platform stack combination based upon the maintenance costs, number of customers, and stability of the customer base.