To my thought we needn't preserve the order in the DataDomain case. It
brings us more flexible model still users given the opportunity to support
it on their own (priority queues can be in handy, I believe). While this
approach delivers additional conveniences if dealing with domains it is
definitely unacceptable in the DataNode case due to strong associations of
nodes with consistent monolite sources of data like databases. Usually one
prefers to mantain a kind of order and certainty during processing the data
with a list of queries when it is tightly related and comes from the same
source.
Andriy.
> Here is the question. When a list of queries comes to a DataDomain, do you
> think it is important to preserve their absolute order, or the only thing
> that matters is the order per DataNode?
>
> If we ignore query ordering across data nodes this gives a potential of
> running individual DataNode query batches in parallel threads. Do you
think
> this might be an important feature? Of course, this places some extra load
> on QueryResultConsumer 's mentioned above....
>
> Or is this the feature used in 0.001% of cases and we should not worry
> about it at all?
>
>
>
> Andrei
>
>
This archive was generated by hypermail 2b30 : Sat Aug 04 2001 - 16:21:24 EDT