>I have a question to the maximum number of nodes. Is there a way to
process
>more nodes or is this planned.
The problem is that DTM is using an int as its internal reference to a
node, and some of those bits are reserved to select between DTM instances.
We can certainly trade off more nodes against fewer documents loadable at a
time -- I think the code's even parameterized so this is a matter of
changing a single field -- but of course that has its own costs.
We might be able to do something clever with cascading two consecutive
blocks of DTM nodes... but I'd be a bit worried about performance effects.
Adding a level of indirection, or moving the which-DTM-does-it-belong-to
into an explicit field on the node, would also have performance costs, and
storage costs as well.
This basically goes in the same category as our observation that DTM makes
source-tree pruning considerably more difficult. It's a known issue, and
we're still considering how to address it.
