On Tue, 2002-03-19 at 07:46, mlw wrote:
> I was thinking about this. There seems to be a consensus that caching means no
> ACID compliance. And everyone seems to think it needs to be limited. We can
> implement a non-ACID cache as a contrib function with some work to the function
> manager.
Until know, I hadn't really thought about it...I just took it for
granted since it was asserted...however, what isn't ACID about the
approach that I offered?
A - Not effected...it's read only and provided directly from the
database, thus, it's still a function of the database. Any change
resulting from atomic changes are notified to the cache, whereby it is
repopulated.
C - Not effected...the database is still responsible for keeping
consistency. The cache is still read only. State is ensured as
invalidation is notified by the database and data set should be returned
consistent by the database or the database is broken.
I - Again, the database is still performing this task and notifies the
cache when updates need to take place. Again, Isolation isn't an issue
because the cache is still read-only.
D - Durability isn't a question either as, again, the database is still
doing this. In the event of cache failure...it would be repopulated
from the database...so it would be as durable as is the database.
Please help me understand.
Thanks,Greg