sharing query result

[Mark Smith]

> Just another thought. I notice that the replicators in the same host can
> share the query results with one another inorder to reduce database
> polling using some kinds of ipc or ithread.

Realistically, yes.

It'd be very nice to have some sort of replication manager that
identifies files that need replicating, and then pass that information
on down to the replication workers.

Even better, setup the UDP transport between the trackers that we've
been wanting since the beginning, and have them elect someone to
'master' the process and pass out tasks to everybody.

The blocking factor will be the actual replication and not figuring out
what needs replicating.  This is actually more feasible given the
architecture that we've got going now, but would still need some work...


--
Mark Smith
junior at danga.com