|
From: Doug C. <cu...@ap...> - 2008-02-15 00:12:47
|
Doug Cutting wrote: > First I'll implement a completely synchronous version, then write the > simulator/tester. Once that's working, we can start designing a > threaded, pseudo-distributed version [ ... ] The synchronous implementation will also be useful as the "model" when testing: each change and query can be made to both the static model and to the "distributed" implementation, comparing the results. The model may sometimes return slightly different results, e.g., when updates are not yet fully synchronized, so testing will log differences from the model rather than fail entirely. Doug |