The reason Infinispan currently uses transactions and forces JTA transactions,
for any clustering mode that requires entity updates, is because the need to
rollback contents if they’re partially applied in the cluster. Without
transactions, we do not have the capabilities to rollback changes if
Y, Ehcache will work clustered both in transactional (JTA), read-only and
read-write just fine (non strict will work, only blows the race wider open
and inconsistent data will be noticed "quicker").
Am trying to think hard what a cache provider "conceptually needs" to
support clustered scenarios. O
Both Ehcache and Infinispan support transactions, but Ehcache doesn't
use this capability when running as Hibernate 2nd level cache, while
Infinispan uses it more as an implementation detail to provide
internal consistency, and not to participate with the application
transaction.
So I think that t
