Thanks Pete, Here is my view to your principles, Ziu, Peter wrote:
Keisuke,
I believe that the most prudent solution needs to take into account the following two principles:
- the repository needs to be able to handle (within the limits of the hardware set) any amount of application contents and associated data for a system (one notional AA).
I fully agree with this and think the protocol choice for bulk data transportation derived from this consideration.
- Something must maintain the notion of the version and integrity of system (via one notional AA), through proper referential integrity and signature.
Yes, something must. However, I think it is arguable that the ACS is the one or we should leave it for the specific implementations. For example, the version numbering model would implicitly include how, who and when will create branches in the versions. I'm not sure if we can safely specify these details as an single standard. -Keisuke