[MarkPNeyer]

http://dl.acm.org/citation.cfm?id=564601

this is the paper in which the theorem was formalized. they talk in terms of 'the initial value' of an atomic object, and its 'subsequent values' - so this only makes sense when the system is _implemented_ in terms of objects which have values that change over time as the result of operations:

--- More formally, let v0 be the initial value of the atomic object. Let c~1 be the prefix of an execution of A in which a single write of a value not equal to v0 occurs in G1, e ---

If requests that come in are expressed as "give me the most recent value of A," and the response comes back with a history of all operations done on A along with timestamps, then you're no longer bound by the CAP theorem because 'an operation was missing' is no longer a _problem_ - its' just an outdated answer.