On Wednesday, 08/28/2002 at 08:55 MST, John Gentilin <[EMAIL PROTECTED]> wrote: > The easiest way to set up the test bed is to configure InstantDB
OK; now to find a copy of InstantDB... > When you say "fragmentary-DTM implementation" .. I'm referring to the fact that -- as you explained -- you are reusing DTM nodes to represent new data, and in streaming mode discarding some nodes as you go, both of which break DTM APIs assumptions. You're counting on understanding DTM usage well enough to write stylesheets which avoid exposing these shortcuts. It's a fairly elegant hack, but it is a hack, and inherently fragile as Xalan continues to evolve. I think we can probably get it working again, but I'm not sure how much longer we can count on it continuing to work. > If there is something I can do to make the implementation more complete please > let me know. As I say, I suspect that the long-term answer -- IF/WHEN we get DTM pruning working on the main source tree -- is going to be to treat this like any other DTM, building a complete and consistant tree model which supports normal traversal and then discarding portions which we've determined will no longer affect the results. That approach may be a bit less efficient (since it does mean generating and discarding nodes rather than moving them) but should be more maintainable. But for the short term, we need to figure out whether we can get the current solution back on the air. ______________________________________ Joe Kesselman / IBM Research
