[vtkusers] streaming functionality information

[tom fogal]

Hi, I was hoping to get some information on streaming that I can't seem
to find anywhere else.

First off is the details on how it all works. From what I understand
currently, it requires a parallel reader. The parallel reader must
chunk the data into 'pieces'. Are these pieces attached to the object
that the reader outputs? Or does some chain of Update/Execute calls
figure out automagically when it needs to overwrite/replace the output
object?

On a related note, how much knowledge of streaming do objects
downstream in the pipeline need to have? Does a filter need to be
written to expect data will come in pieces, such that the filter needs
to work on chunk 1, update chunk 2, then work on chunk 2, update chunk
3, etc. Also how far down the pipeline must one go before streaming
becomes transparent?

My final question relates to the current status of streaming. We are
investigating writing a streaming reader, however I'm looking through
some downstream filters and not seeing much in the way of handling
'pieces' of data. The objects I am seeing it in seem to relate mostly
to merging, converting, or somehow rearranging data sets though, so I'm
hoping that filters which only \emph{use} the input data set, not
modify it, don't actually need to be written with streaming in mind.
I'm also looking at VTK4.4; CVS might be farther along.

Thanks for any light you could shed on the subject!

-tom