[rabbitmq-discuss] Fwd: question on the faq

[Alexis Richardson]

Carl,

On Mon, Jan 5, 2009 at 4:12 PM, Carl Trieloff <cctrieloff at redhat.com> wrote:
>
>> Only if the spec says that it should be, which it doesn't.
>
> that is ridiculous.

The spec could always be improved - please suggest language to prevent
interpretations that you consider to be ridiculous.



> I don't know of any transaction impl that is not
> ACID,

Well there is not one 'ACID' and there is not one 'transaction'
definition or impl, which is why there are so many long and beardy
books about the subject.

I would argue that eventual conistency (Dynamo, SimpleDB) is not ACID
for some classes of observation - roughly speaking, where you are
willing to let processes observe the system from different points.

I would also argue that compensating transactions weaken ACIDity
significantly which is why WS-* has both WS-AtomicTransaction and the
'business transaction' stuff - see my link from an earlier email,
referencing Sanjiva's comment on TSS.

alexis


> if it is not ACID what is the point? Differentiating between 1PC
> (local) and 2PC is meaningful, but both are ACID, that is the point on a
> transaction. If the txn is not ACID, why not just acknowledge the
> message, there is to difference in the failure cases between a non ACID
> txn and an ack  then.
>
> Carl.
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>